AI model-based rental space matching platform system

Abstract

We introduce an AI model-based rental space matching platform system that improves the accuracy of matching tenants optimized for rental spaces by performing commercial area analysis and business feasibility analysis by industry based on rental space information entered by building owners, and automatically determining the priority of eligible tenants using an AI model. The AI ​​model-based rental space matching platform system includes: a matching unit that receives rental space information from a building owner's terminal and provides information on businesses available for entry and information on interior design companies; a business selection unit that selects businesses for entry based on rental space information transmitted from the matching unit, determines entry priority, and performs matching processing; a design input unit that receives interior design data from the representative terminal of an interior design company selected through the matching unit; a model generation unit that inputs the interior design data transmitted from the design input unit into a pre-trained AI model to generate a 3D virtual interior model including the interior and exterior of the rental space; a visualization unit that provides the 3D virtual interior model generated from the model generation unit to the building owner's terminal or a customer's terminal, receives modification requests, and transmits them to the model generation unit; and a data management unit that stores matching result data transmitted from the business selection unit, 3D virtual interior model data generated from the model generation unit, and modification history data transmitted from the visualization unit, and manages entry history and interior progress status for each rental space.

Description

Technology Field

[0001] The present invention relates to an artificial intelligence model-based rental space matching platform system, and more specifically, to an artificial intelligence model-based rental space matching platform system that selects eligible tenants and determines priorities based on rental space information input by a building owner, and converts interior design data into a 3D virtual interior model through an artificial intelligence model and provides it to the building owner or customer, thereby providing integrated support from matching tenants for rental spaces to interior design and visualization. Background Technology

[0002] In the recent commercial real estate rental market, there is a problem in that the process of building owners finding suitable tenants for rental spaces and planning interiors tailored to the characteristics of those businesses is very complex and time-consuming.

[0003] Traditional methods of matching rental spaces primarily involved building owners directly searching for prospective tenants or passive matching through real estate agencies. This approach failed to systematically analyze factors such as location, commercial district analysis, and business feasibility by industry, resulting in high business failure rates after occupancy and causing economic losses for both building owners and tenants.

[0004] Furthermore, even during the interior design phase after a tenant is decided, it is difficult to accurately predict the final interior outcome in advance using only 2D floor plans; consequently, building owners or tenants frequently become dissatisfied with results that differ from their expectations after construction. In particular, when requests for interior modifications arise during the construction phase, additional costs and time are incurred, which lowers the overall efficiency of the project.

[0005] Recently, with the advancement of 3D modeling and artificial intelligence technologies, attempts are being made to utilize them to visualize interior designs and analyze business feasibility; however, there is still a lack of systems that provide integrated support for the entire process, from matching rental spaces to interior design, 3D model creation, modification, and approval.

[0006] Therefore, there is an urgent need to develop an integrated platform system that can maximize the efficiency of rental space matching and interior design by analyzing the characteristics of rental spaces to select optimal tenants, using artificial intelligence models to automatically convert interior design drawings into 3D virtual models, and providing them to building owners and customers in a visualized form. Prior art literature

[0007] KR 10-2023-0190450 The problem to be solved

[0008] The present invention has been devised to solve the problems of the prior art described above, and provides an AI model-based rental space matching platform system that improves the accuracy of matching tenants optimized for rental spaces by performing commercial area analysis and business feasibility analysis by industry based on rental space information input by a building owner, and automatically determining the priority of eligible tenants using an AI model.

[0009] In addition, the present invention aims to support decision-making during the interior design stage and reduce post-construction modification costs by inputting 2D interior design data provided by an interior design company into a pre-trained artificial intelligence model to automatically generate a 3D virtual interior model and providing it to a terminal of a building owner or customer so that the interior results can be checked in advance before actual construction.

[0010] In addition, the present invention aims to derive optimal interior design patterns for each rental space type by receiving a request to modify a 3D virtual interior model from a building owner or customer, modifying the model in real time, and continuously improving the performance of the artificial intelligence model by utilizing the modification history data as training data.

[0011] In addition, the present invention aims to efficiently manage the entire process from matching rental spaces to completing interiors by systematically storing and managing matching result data by rental space, 3D virtual interior model data, and modification history data, and tracking the progress status of the interiors step by step and providing it to the building owner. means of solving the problem

[0012] An artificial intelligence model-based rental space matching platform system according to an embodiment of the present invention for solving the above problem comprises: a matching unit that receives rental space information from a building owner's terminal and provides information on businesses eligible for entry and information on interior design companies; a business selection unit that selects businesses eligible for entry based on rental space information transmitted from the matching unit, determines entry priority, and performs matching processing; a design input unit that receives interior design drawing data from a representative terminal of an interior design company selected through the matching unit; a model generation unit that inputs the interior design drawing data transmitted from the design input unit into a pre-trained artificial intelligence model to generate a three-dimensional virtual interior model including the interior and exterior of the rental space; and a visualization unit that provides the three-dimensional virtual interior model generated from the model generation unit to the building owner's terminal or a customer's terminal, receives a modification request, and transmits it to the model generation unit. and includes a data management unit that stores matching result data transmitted from the above-mentioned vendor selection unit, 3D virtual interior model data generated from the above-mentioned model generation unit, and modification history data transmitted from the above-mentioned visualization unit, and manages the entry history and interior progress status for each leased space.

[0013] The above matching unit comprises: an information input unit that receives rental space information including location information, area information, number of floors, floor height information, actual measurement dimensions, window location information, and rent information from the building owner's terminal, and converts the received rental space information into a structured data format; a response providing unit that provides a list of prospective tenants transmitted from the company selection unit to the building owner's terminal and receives company selection information or an interior company request signal from the building owner's terminal; and an interior linkage unit that constructs an interior company database storing identification information, contact information, and specialized field information for each interior company, generates a list of interior companies corresponding to the rental space by linking with the constructed interior company database when an interior company request signal is received through the response providing unit, provides the generated list of interior companies to the building owner's terminal, and receives the selected interior company identification information and contact information from the building owner's terminal. and a data transmission unit that transmits structured rental space information converted from the information input unit to the company selection unit and the model generation unit, transmits company selection information entered through the response provision unit to the company selection unit, and transmits interior company identification information and contact information entered through the interior linkage unit to the design input unit;It includes, and the business selection unit receives structured rental space information transmitted from the data transmission unit, and performs API communication with an external commercial area analysis system based on location information included in the received rental space information to query the number of pedestrian traffic, the number of surrounding competitors, and the distance from a public transportation station at the corresponding location; assigns 20 points if the queried pedestrian traffic is 10,000 or more, 10 points if it is 5,000 or more but less than 10,000, and 0 points if it is less than 5,000; assigns 20 points if the number of surrounding competitors is less than 3, 10 points if it is 3 or more but less than 5, and 0 points if it is 5 or more; assigns 20 points if the distance from a public transportation station is within 300m, 10 points if it is between 300m and 500m, and 0 points if it is over 500m; calculates a location score by summing the assigned scores and dividing by 60 to normalize the result to a range of 0 to 100, and together with the area information and rent information included in the rental space information, for each rental space A spatial analysis unit that generates characteristic data; a tenant database that constructs a database of tenants that stores industry information, average sales information, investment cost information, preferred location information, preferred area range information, history of past successful entry, and history of past failed entry by tenant, retrieves information by tenant stored in the constructed tenant database, extracts tenants as eligible tenants by comparing the location of the leased space with characteristic data by leased space transmitted from the spatial analysis unit, and generates tenant condition data by tenant that includes the extracted industry information, average sales information, investment cost information, number of past entry attempts, and number of past successful entry.A priority determination unit that combines characteristic data for each rental space transmitted from the space analysis unit and entry condition data for each company transmitted from the company inquiry unit, calculates a business feasibility score for each company using a pre-trained first artificial intelligence model, calculates an entry priority weight based on the calculated business feasibility score, the location score transmitted from the space analysis unit, and the number of past entry attempts and past entry successes transmitted from the company inquiry unit, and sorts the list of target companies for entry according to the calculated entry priority weight; a matching execution unit that transmits the list of target companies sorted by the priority determination unit to the response provision unit, receives company selection information transmitted from the data transmission unit, transmits matching request data including the rental space information, building owner contact information, and matching request time information to the headquarters system of the corresponding company, and transmits matching result data to the data management unit after the matching is completed; and includes a vendor data storage member that manages the aforementioned vendor database, and a vendor management unit comprising a vendor evaluation member that calculates a vendor success rate based on the past number of vendor entry attempts and past number of successful entry attempts of each vendor transmitted from the vendor inquiry unit, wherein if the past number of vendor entry attempts is 0, a basic vendor success rate of 50 is assigned, and if the past number of vendor entry attempts is 1 or more, the vendor success rate is calculated as a percentage by multiplying the value obtained by dividing the past number of successful entry attempts by the past number of vendor entry attempts by 100, and transmits the calculated vendor success rate to the aforementioned priority determination unit; wherein the vendor entry priority weight is calculated by the following [Mathematical Formula 1];

[0014] [Mathematical Formula 1]

[0015] W = α × B + β × L + γ × S

[0016] (Here, W is the store entry priority weight, B is the business feasibility score calculated by the priority determination unit (normalized to a value between 0 and 100), L is the location score calculated by the spatial analysis unit (normalized to a value between 0 and 100), S is the store entry success rate calculated by the business evaluation unit (percentage, between 0 and 100), α, β, and γ represent the weighting coefficients for the business feasibility score, location score, and store entry success rate, respectively, and α, β, and γ are each values ​​between 0 and 1, α + β + γ = 1, and a larger W value indicates a higher store entry priority.)

[0017] The above-mentioned first artificial intelligence model learns past store entry success data, average sales data by industry, commercial area analysis data by region, and profitability data relative to rent to improve the accuracy of calculating the business feasibility score; the above-mentioned priority determination unit calculates the expected monthly sales, expected investment recovery period, and expected operating profit margin for each company through the above-mentioned first artificial intelligence model, assigns a higher value as the expected monthly sales are higher, the expected investment recovery period is shorter, and the expected operating profit margin is higher, thereby normalizing the business feasibility score to a range of 0 to 100; and the above-mentioned priority determination unit excludes companies whose calculated store entry priority weight is less than a preset minimum priority standard value from the list of eligible companies for store entry.

[0018] The system further includes an industry database storing average conversion rates, average transaction value, average cost ratio, standard monthly sales, standard payback period, and standard monthly net profit for each industry sector, including restaurants, cafes, unmanned stores, convenience stores, gyms, and academies; wherein the design input unit comprises: a communication connection unit that performs a communication connection with a representative terminal of an interior design company based on interior design company identification information and contact information transmitted from the data transmission unit; and a design data reception unit that receives interior design drawing data including at least one of a 2D floor plan, elevation, and cross-section from the representative terminal of the interior design company through the communication connection unit, and extracts wall information, furniture placement information, finishing material information, and lighting information from the received interior design drawing data. and a data conversion unit that converts wall information, furniture placement information, finishing material information, and lighting information extracted from the design data receiving unit into a structured design data format and transmits the converted structured design data to the model generation unit; wherein the model generation unit receives the structured design data transmitted from the data conversion unit and receives actual measurement dimension information, floor height information, and window location information of the rental space transmitted from the data transmission unit, generates a basic space model in the form of a three-dimensional rectangular prism having width, length, and height based on the received actual measurement dimension information and floor height information, and completes the three-dimensional basic space model by forming an opening at the corresponding location according to the received window location information; An internal modeling unit that generates an internal space partitioning model by applying wall information included in structured design data transmitted from the data conversion unit to a 3D space basic model generated from the spatial modeling unit to place walls that partition the internal space, places furniture objects in 3D coordinates by applying furniture placement information included in the structured design data to the generated internal space partitioning model, and generates a furniture placement model by performing rotation and scaling for each placed furniture object;An external modeling unit that applies external finishing material information, signboard information, and entrance information included in the structured design data transmitted from the data conversion unit to the exterior wall area of ​​the 3D spatial basic model generated from the spatial modeling unit, generates a signboard model based on the location, size, shape, and color of the signboard included in the signboard information, and generates an exterior model by placing the generated signboard model at a designated location on the exterior wall; a rendering unit that performs texture mapping according to the finishing material information included in the structured design data transmitted from the data conversion unit on the furniture placement model generated from the interior modeling unit and the exterior model generated from the external modeling unit, places a light source in a 3D space based on lighting type, lighting location, lighting brightness, and color temperature information according to the lighting information included in the structured design data, calculates the reflection, shadow, and contrast of light generated from the placed light source to perform photorealistic rendering to generate a final 3D virtual interior model, and generates image data captured from multiple viewpoints and 360-degree panoramic image data of the generated final 3D virtual interior model. AI processing unit comprising: a model storage member that stores a pre-trained second artificial intelligence model; and a model operation member that inputs structured design data transmitted from the data conversion unit into the second artificial intelligence model to generate 3D coordinate information, object placement information, and texture information, and transmits the generated 3D coordinate information, object placement information, and texture information to the spatial modeling unit, the internal modeling unit, the external modeling unit, and the rendering unit, respectively, wherein if there is information explicitly specified in the structured design data, such information is applied first, and if there is no explicit information, information generated by the second artificial intelligence model is used as auxiliary information;It includes, wherein the second artificial intelligence model is a generative artificial intelligence model that has learned past interior design drawing data and 3D model data corresponding to the design drawing, and is trained to receive a 2D design drawing image as input and output 3D spatial coordinates, the location and size of an object, and material and color information; the rendering unit transmits the generated final 3D virtual interior model data, multiple viewpoint image data, and 360-degree panoramic image data to the visualization unit; and the priority determination unit includes a commercial area data storage unit that stores the number of floating population, the number of surrounding competitors, and distance information from public transportation stations included in the characteristic data for each rental space transmitted from the spatial analysis unit; and a sales prediction unit that queries the average conversion rate and average transaction value of the corresponding industry from the industry database based on industry information included in the entry condition data for each business transmitted from the business inquiry unit, calculates the estimated number of customers by multiplying the number of floating population stored in the commercial area data storage unit by the queried average conversion rate, and calculates the estimated monthly sales by multiplying the calculated estimated number of customers by the queried average transaction value. A profitability analysis unit that calculates monthly operating costs including labor costs, material costs, management costs, and marketing costs by multiplying the estimated monthly sales by the average cost ratio by industry retrieved from the industry database, based on investment cost information included in the store entry condition data for each company transmitted from the above company inquiry unit and the estimated monthly sales calculated from the above sales forecast unit; calculates monthly net profit by subtracting the monthly rent entered through the above information input unit and the calculated monthly operating costs from the estimated monthly sales; and calculates the investment recovery period by dividing the initial investment cost included in the above investment cost information by the above monthly net profit;and a score calculation member that combines the expected monthly sales calculated from the sales forecasting member, the investment recovery period calculated from the profitability analysis member, and the monthly net profit, calculates a business feasibility score according to the following [Mathematical Formula 2], adjusts the calculated business feasibility score to 100 if it exceeds 100 and to 0 if it is less than 0 to generate a final business feasibility score, and transmits the generated final business feasibility score to the priority determination member;

[0019] [Mathematical Formula 2]

[0020] B = δ × (R / R_ref) × 100 + ε × (P_ref / P) × 100 + ζ × (M / M_ref) × 100

[0021] (Here, B is the business feasibility score, R is the expected monthly revenue calculated from the sales forecasting component, R_ref is the standard monthly revenue of the corresponding industry stored in the industry database, P is the payback period (in months) calculated from the profitability analysis component, P_ref is the standard payback period (in months) of the corresponding industry stored in the industry database, M is the monthly net profit calculated from the profitability analysis component, M_ref is the standard monthly net profit of the corresponding industry stored in the industry database, δ, ε, and ζ represent weighting coefficients for the expected monthly revenue, payback period, and monthly net profit, respectively; δ, ε, and ζ are values ​​between 0 and 1, respectively, δ + ε + ζ = 1, and a higher final business feasibility score, adjusted from 0 to 100, indicates higher business feasibility.)

[0022] The above priority determination unit calculates the entry priority weight W using the final business feasibility score transmitted from the score calculation unit as the input value B of [Mathematical Formula 1], and the above priority determination unit excludes companies from the list of eligible companies for entry if the final business feasibility score transmitted from the score calculation unit is less than the preset minimum business feasibility standard score.

[0023] The visualization unit comprises: a model receiving unit that receives final 3D virtual interior model data, a plurality of viewpoint image data, and 360-degree panoramic image data transmitted from the rendering unit, and converts the received data into a format transmittable to the building owner's terminal or the customer's terminal; a model providing unit that transmits the 3D virtual interior model data converted from the model receiving unit to the building owner's terminal or the customer's terminal, causes the 3D virtual interior model to be displayed on the display unit of the building owner's terminal or the customer's terminal, and provides an interactive viewer function to enable rotation, zooming in, zooming out, and viewpoint changes for the displayed 3D virtual interior model; and a modification request input unit that receives modification request information including at least one of a request to change furniture placement, a request to change finishing materials, a request to change lighting, and a request to change color from the building owner's terminal or the customer's terminal, and analyzes the input modification request information to extract information on the object to be modified and information on the modification content. and a modification processing unit that transmits modification target object information and modification content information extracted from the modification request input unit to the model generation unit, transmits the modification request information to the interior modeling unit if the modification request information is a request for furniture arrangement change, transmits the modification request information to the exterior modeling unit if the modification request information is a request for exterior finishing material change or signboard change, transmits the modification request information to the rendering unit if the modification request information is a request for finishing material change, lighting change, or color change, receives the modified final 3D virtual interior model data from the rendering unit and provides it to the building owner's terminal or the customer's terminal, and generates modification history data including modification request time information, modification content information, and modification count information and transmits it to the data management unit;The model providing unit rotates the viewpoint of the 3D virtual interior model up, down, left, and right in response to a touch on the building owner's terminal or the customer's terminal, and performs zooming in or out in response to a pinch gesture or scroll input; the model providing unit provides a virtual reality mode using 360-degree panoramic image data transmitted from the model receiving unit, and enables the user to freely look around the interior of the 3D virtual interior model through a display in the virtual reality mode; the modification request input unit receives input from the user selecting a specific object on the 3D virtual interior model displayed on the building owner's terminal or the customer's terminal, and receives at least one modification command among moving, rotating, replacing, or deleting the selected object; the modification processing unit transmits identification information of the furniture object to be modified and location coordinate information to be changed to the interior modeling unit if the modification request information is a request to change the finishing material, and transmits area information of the wall or floor to be modified and texture information of the finishing material to be changed to the rendering unit if the modification request information is a request to change the finishing material; and the modification processing unit [requires] three or more modification requests for the same 3D virtual interior model. In the case of repetition, the pattern of the corresponding modification request is analyzed to extract information on frequently modified objects and modification directions, and the extracted information is transmitted to the AI ​​processing unit to be used as training data for the second artificial intelligence model; the modification processing unit includes modification request time information, modification target object information, state information before modification, state information after modification, and modification requester information in the modification history data and transmits it to the data management unit.

[0024] The above data management unit comprises: a matching data storage unit that receives matching result data transmitted from the matching execution unit, extracts rental space identification information, matched interior company identification information, and matching completion time information included in the matching result data, and stores them in an entry history database for each rental space; a model data storage unit that receives 3D virtual interior model data transmitted from the rendering unit, tags the rental space identification information, interior company identification information, and model creation time information on the 3D virtual interior model data, and stores it in an interior model database; a modification history storage unit that receives modification history data transmitted from the modification processing unit, analyzes the modification request time information, modification target object information, and modification content information included in the modification history data, and stores them in an entry history database for each rental space; and a progress status management unit that calculates the number of matching attempts, the number of matching successes, and the average matching time required for each rental space based on the entry history data stored in the matching data storage unit, determines the interior progress stage for each rental space based on the 3D virtual interior model creation history data stored in the model data storage unit, and provides the determined interior progress stage information to the building owner's terminal. and a learning data generation unit that analyzes modification history data stored in the modification history storage unit to statistically process frequently modified items by rental space type, inputs the statistically processed data into a pre-trained third artificial intelligence model to derive optimal interior design patterns by rental space type, and transmits the derived optimal interior design patterns to the AI ​​processing unit to be used as training data for the second artificial intelligence model;It includes, wherein the progress status management unit divides the interior progress stage into five stages: matching completion stage, design progress stage, model creation stage, modification progress stage, and final approval stage; wherein the progress status management unit determines the matching completion stage when matching result data is stored in the matching data storage unit, determines the design progress stage when it receives a signal that the interior design drawing data reception is complete from the design data reception unit, determines the model creation stage when 3D virtual interior model data is stored in the model data storage unit, determines the modification progress stage when modification history data is stored in the modification history storage unit, and determines the final approval stage when it receives a final approval signal from the building owner's terminal; wherein the learning data generation unit classifies objects with two or more modifications among the modification history data stored in the modification history storage unit as frequently modified objects, extracts state information after the final modification for the frequently modified objects, and generates preferred design data; wherein the learning data generation unit classifies the rental space type into restaurant, cafe, unmanned store, office, and storage, and inputs the preferred design data collected for each type into the third artificial intelligence model. It derives optimal furniture placement patterns, optimal color combinations, and optimal lighting placements by type, and the third artificial intelligence model learns past modification history data, finally approved 3D virtual interior model data, and rental space type information to improve the accuracy of deriving optimal interior design patterns by rental space type. Effects of the invention

[0025] According to the present invention, based on information such as the location, area, and rent of a rental space, the characteristics of the rental space can be objectively and quantitatively evaluated by automatically analyzing the floating population, competitors, and public transportation accessibility in conjunction with a commercial area analysis system and calculating a location score based on this.

[0026] In addition, by utilizing the first artificial intelligence model to calculate business feasibility scores by industry and automatically calculating priority weights for store entry that combine location scores and success rates, and sorting the list of potential tenants, it is possible to recommend optimal tenants to building owners and reduce the business failure rate after entry.

[0027] In addition, by inputting 2D design data provided by an interior design company into a second artificial intelligence model to automatically generate a 3D virtual interior model, construction results can be visualized in advance during the interior design phase, and decision-making by building owners and customers can be supported.

[0028] In addition, by providing the generated 3D virtual interior model to the building owner's or customer's terminal and offering an interactive viewer function and a virtual reality mode, users can view the interior from various perspectives and input modification requests in real time to have them reflected immediately, thereby maximizing the efficiency of the interior design.

[0029] In addition, by analyzing modification history data to derive optimal interior design patterns for each rental space type through a third AI model and utilizing this as training data for a second AI model, it is possible to continuously improve the performance of the AI ​​model and generate more accurate 3D models for similar projects in the future.

[0030] In addition, by systematically storing matching result data, 3D virtual interior model data, and modification history data, and managing the interior progress stages by dividing them into five steps, the building owner can monitor the project's progress in real time and improve the transparency and management efficiency of the entire process. Brief explanation of the drawing

[0031] FIG. 1 is a configuration diagram of an artificial intelligence model-based rental space matching platform system according to one embodiment of the present invention. FIG. 2 is a block diagram of a matching unit according to one embodiment of the present invention. FIG. 3 is a block diagram of a company selection unit according to one embodiment of the present invention. FIG. 4 is a block diagram of a design input unit according to one embodiment of the present invention. FIG. 5 is a block diagram of a model generation unit according to one embodiment of the present invention. FIG. 6 is a block diagram of a visualization unit according to one embodiment of the present invention. FIG. 7 is a block diagram of a data management unit according to one embodiment of the present invention. Figure 8 is a drawing illustrating the process of inputting a design drawing through the representative terminal of an interior design company. Figure 9 is a drawing illustrating how an interior model created on a customer's terminal and a building owner's terminal is displayed. Figure 10 is a drawing illustrating the parts to be modified in an interior model created through the customer's terminal and the building owner's terminal. FIG. 11 is a drawing illustrating how an interior model before and after modification is displayed through a customer's terminal and a building owner's terminal according to an embodiment of the present invention. Specific details for implementing the invention

[0032] Hereinafter, embodiments are described in detail with reference to the attached drawings. However, various modifications may be made to the embodiments, and thus the scope of the patent application is not limited or restricted by these embodiments. It should be understood that all modifications, equivalents, and substitutions to the embodiments are included within the scope of the rights.

[0033] Specific structural or functional descriptions of the embodiments are disclosed for illustrative purposes only and may be modified and implemented in various forms. Accordingly, the embodiments are not limited to the specific disclosed forms, and the scope of this specification includes modifications, equivalents, or substitutions that fall within the technical concept.

[0034] Terms such as "first" or "second" may be used to describe various components, but these terms should be interpreted solely for the purpose of distinguishing one component from another. For example, the first component may be named the second component, and similarly, the second component may be named the first component.

[0035] When it is stated that a component is "connected" to another component, it should be understood that it may be directly connected to or coupled with that other component, or that there may be other components in between.

[0036] The terms used in the embodiments are for illustrative purposes only and should not be interpreted as intended to be limiting. Singular expressions include plural expressions unless the context clearly indicates otherwise. In this specification, terms such as "comprising" or "having" are intended to indicate the existence of the features, numbers, steps, actions, components, parts, or combinations thereof described in the specification, and should be understood as not precluding the existence or addition of one or more other features, numbers, steps, actions, components, parts, or combinations thereof.

[0037] Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as generally understood by those skilled in the art to which the embodiments pertain. Terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant technology, and should not be interpreted in an ideal or overly formal sense unless explicitly defined in this application.

[0038] In addition, when describing with reference to the attached drawings, identical components are assigned the same reference numeral regardless of drawing symbols, and redundant descriptions thereof are omitted. In describing the embodiments, if it is determined that a detailed description of related prior art could unnecessarily obscure the essence of the embodiments, such detailed description is omitted.

[0039] The advantages and features of the present invention and the methods for achieving them will become clear by referring to the embodiments described below in detail together with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below but may be implemented in various different forms. These embodiments are provided merely to ensure that the disclosure of the present invention is complete and to fully inform those skilled in the art of the scope of the invention, and the present invention is defined only by the scope of the claims.

[0040] In the embodiments of the present invention, unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as generally understood by those skilled in the art to which the present invention pertains. Terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant technology, and should not be interpreted in an ideal or overly formal sense unless explicitly defined in the embodiments of the present invention.

[0041] The shapes, sizes, ratios, angles, numbers, etc. disclosed in the drawings for explaining embodiments of the present invention are exemplary, and therefore the present invention is not limited to the depicted details. Furthermore, in describing the present invention, if it is determined that a detailed description of related known technology may unnecessarily obscure the essence of the present invention, such detailed description is omitted. Where terms such as "includes," "has," or "is made up" are used in this specification, other parts may be added unless "only" is used. Where a component is expressed in the singular, it includes cases where it includes the plural unless specifically stated otherwise.

[0042] In interpreting the components, they are interpreted to include a margin of error even in the absence of a separate explicit statement.

[0043] In the case of describing a positional relationship, for example, when the positional relationship between two parts is described using expressions such as 'on,' 'upper,' 'lower,' or 'next to,' one or more other parts may be located between the two parts unless 'immediately' or 'directly' is used.

[0044] When elements or layers are referred to as "on" another element or layer, this includes cases where another layer or element is placed directly on top of or in between. Throughout the specification, the same reference numerals refer to the same components.

[0045] The size and thickness of each component shown in the drawings are illustrated for convenience of explanation, and the present invention is not necessarily limited to the size and thickness of the illustrated components.

[0046] The features of each of the various embodiments of the present invention may be combined or combined with one another, either partially or wholly, and as will be fully understood by those skilled in the art, various technical interlocking and operation are possible, and each embodiment may be implemented independently of one another or together in an interlocking relationship.

[0048] FIG. 1 is a configuration diagram of an artificial intelligence model-based rental space matching platform system according to an embodiment of the present invention. FIG. 2 is a block diagram of a matching unit according to an embodiment of the present invention. FIG. 3 is a block diagram of a vendor selection unit according to an embodiment of the present invention. FIG. 4 is a block diagram of a design input unit according to an embodiment of the present invention. FIG. 5 is a block diagram of a model generation unit according to an embodiment of the present invention. FIG. 6 is a block diagram of a visualization unit according to an embodiment of the present invention. FIG. 7 is a block diagram of a data management unit according to an embodiment of the present invention. FIG. 8 is a diagram illustrating the input of a design drawing through a representative terminal of an interior design company. FIG. 9 is a diagram illustrating the display of an interior model created on a customer's terminal and a building owner's terminal. FIG. 10 is a diagram illustrating the display of parts to be modified in the interior model created through a customer's terminal and a building owner's terminal. FIG. 11 is a diagram illustrating the display of the interior model before and after modification through a customer's terminal and a building owner's terminal according to an embodiment of the present invention.

[0049] Hereinafter, with reference to FIGS. 1 to 11, an artificial intelligence model-based rental space matching platform system () according to an embodiment of the present invention will be described in more detail.

[0050] An artificial intelligence model-based rental space matching platform system (100) according to one embodiment of the present invention includes a matching unit (110), a vendor selection unit (120), a design input unit (130), a model generation unit (140), a visualization unit (150), and a data management unit (160).

[0051] The AI ​​model-based rental space matching platform system (100) receives rental space information from the building owner's terminal (P1), selects businesses that can move in, matches the selected businesses with the building owner, and converts design data provided by the interior design company into a 3D virtual interior model through an AI model and provides it to the building owner's terminal (P1) or the customer's terminal (P2). The AI ​​model-based rental space matching platform system (100) comprehensively manages the entire process from matching businesses for rental spaces to interior design and visualization, and maximizes the efficiency of the entire project through organic data linkage between each unit.

[0052] The building owner's terminal (P1) may be any one of a smartphone, tablet PC, laptop, or desktop computer, and is connected to an artificial intelligence model-based rental space matching platform system (100) via a wireless communication network or a wired communication network. The building owner's terminal (P1) includes an input interface for inputting rental space information and a display unit (not shown) for displaying a list of tenants and a 3D virtual interior model.

[0053] The customer's terminal (P2) is a terminal used by the representative or person in charge of the prospective tenant, and, like the building owner's terminal (P1), it can be any one of a smartphone, tablet PC, laptop, or desktop computer. The customer's terminal (P2) can receive and display a 3D virtual interior model from the artificial intelligence model-based rental space matching platform system (100) and input a request for modification.

[0054] The representative terminal (P3) of the interior design company is a terminal used by the design manager of the interior design company and transmits interior design drawing data to the artificial intelligence model-based rental space matching platform system (100). The representative terminal (P3) of the interior design company provides a function to upload design drawing files created by linking with CAD software and 3D modeling software.

[0055] The matching unit (110) receives rental space information from the building owner's terminal (P1) and provides information on businesses available for entry and interior design businesses. The matching unit (110) converts the rental space information into a structured data format and transmits it to the business selection unit (120) and the model generation unit (140), and mediates the flow of information between the building owner's terminal (P1) and the business selection unit (120). The matching unit (110) acts as an input gateway for the entire system and converts raw data input from the building owner into a standardized format that each unit can process, thereby enabling subsequent units to operate based on a consistent data structure.

[0056] The matching unit (110) includes an information input unit (111), a response providing unit (112), an interior linkage unit (113), and a data transmission unit (114).

[0057] The information input unit (111) receives rental space information including location information, area information, number of floors, floor height information, actual measurement dimensions, window location information, and rent information from the building owner's terminal (P1). The information input unit (111) receives each information item through an input form displayed on the building owner's terminal (P1) and verifies whether the format of the entered information is correct.

[0058] The information input unit (111) converts the received rental space information into a structured data format. The information input unit (111) converts location information into latitude and longitude coordinates, standardizes area information into square meter units, and converts actual measurement dimension information into three-dimensional coordinate data of width, length, and height. The information input unit (111) converts window location information into relative coordinates based on the wall surface and stores it.

[0059] The information input unit (111) receives rental information as a monthly amount and stores the deposit information and monthly rental information separately. The information input unit (111) performs validation on all entered information to check whether required items are missing and whether the entered figures are within a reasonable range.

[0060] The information input unit (111) transmits the converted structured rental space information to the data transmission unit (114), and at this time, complies with the data transmission protocol through the connection interface with the data transmission unit (114). A data integrity verification mechanism operates between the information input unit (111) and the data transmission unit (114), and the data transmission unit (114) calculates the checksum of the received data and compares it with the checksum transmitted by the information input unit (111) to confirm that no data damage occurred during the transmission process. The information input unit (111) assigns a unique identification number to each rental space so that the corresponding rental space can be identified in all subsequent processing processes, and this unique identification number is consistently used in all processing processes of the vendor selection unit (120), model creation unit (140), visualization unit (150), and data management unit (160) through the data transmission unit (114) to ensure data traceability.

[0061] The response providing unit (112) provides the list of prospective tenants transmitted from the tenant selection unit (120) to the building owner's terminal (P1). The response providing unit (112) maintains a real-time data linkage channel with the matching execution unit (124) of the tenant selection unit (120) and receives the information as soon as the matching execution unit (124) generates the list of prospective tenants. The response providing unit (112) receives the list of prospective tenants sorted according to the priority weighting of the tenants and provides the industry information, average sales information, investment cost information, business feasibility score, and tenant success rate of each tenant.

[0062] The response providing unit (112) displays a list of prospective tenants in the form of a table or a card on the display unit of the building owner's terminal (P1). The response providing unit (112) provides an interface for checking detailed information about each tenant, and when the building owner selects a specific tenant, it additionally displays the tenant's past tenancy history, business performance, and brand information.

[0063] The response providing unit (112) receives company selection information or an interior company request signal from the building owner's terminal (P1). When the response providing unit (112) detects input from the building owner selecting a specific company from the list of eligible companies, it extracts identification information of the selected company and generates it as company selection information. When the building owner clicks a button requesting a connection to an interior company, the response providing unit (112) generates an interior company request signal.

[0064] The response providing unit (112) transmits the input company selection information to the data transmission unit (114) and transmits the interior company request signal to the interior linkage unit (113). The response providing unit (112) provides a function that allows multiple companies to be selected simultaneously according to the building owner's selection, and in this case, the identification information of all selected companies is included in the company selection information. An event-based communication method is used in the connection between the response providing unit (112) and the data transmission unit (114), and the data transmission unit (114) is notified immediately when the building owner's selection event occurs, enabling real-time processing.

[0065] The interior linkage unit (113) establishes an interior company database (DB1) in which identification information, contact information, and specialized field information for each interior company are stored. The interior company database (DB1) stores identification information including the business registration number, company name, representative's name, phone number, email address, and office address of each interior company.

[0066] The interior design company database (DB1) stores information on each interior design company's area of ​​expertise, indicating which field—restaurant interiors, cafe interiors, office interiors, or commercial space interiors—they specialize in. The interior design company database (DB1) also stores each company's past construction track record, customer satisfaction ratings, average construction period, and average construction cost.

[0067] When an interior design company request signal is input through the response providing unit (112), the interior design linkage unit (113) links with the established interior design company database (DB1) to generate a list of interior design companies corresponding to the rental space. Based on the area information, floor number information, and expected usage information included in the rental space information transmitted from the information input unit (111), the interior design linkage unit (113) filters interior design companies with specialized fields suitable for the rental space.

[0068] The interior linkage unit (113) generates a list of interior companies by sorting the filtered interior companies in order of customer satisfaction evaluation scores. The interior linkage unit (113) includes images of representative construction cases, average construction costs, and average construction periods for each company in the generated list of interior companies.

[0069] The interior linkage unit (113) provides the generated list of interior companies to the building owner's terminal (P1). The interior linkage unit (113) displays the list of interior companies in the form of cards on the display unit of the building owner's terminal (P1), and each card displays the company name, field of expertise, customer satisfaction score, and an image of a representative construction case.

[0070] The interior linkage unit (113) receives selected interior company identification information and contact information from the building owner's terminal (P1). When the interior linkage unit (113) detects input from the building owner selecting a specific company from the list of interior companies, it retrieves the identification information and contact information of the selected company from the interior company database (DB1).

[0071] The interior linkage unit (113) transmits the extracted interior company identification information and contact information to the data transmission unit (114), and simultaneously sends a preliminary notification so that the design input unit (130) can prepare a communication connection with the interior company. The interior linkage unit (113) sends a matching notification message to the selected interior company to notify it that there is a request for interior design of the rental space, and this notification message includes basic information about the rental space and the building owner's contact information so that the interior company can immediately identify the project.

[0072] The data transmission unit (114) transmits the structured rental space information converted from the information input unit (111) to the company selection unit (120) and the model creation unit (140). The data transmission unit (114) converts the rental space information into structured data in JSON or XML format and transmits it through a network, and at this time, adjusts the data format to match the data reception interface specifications of each unit.

[0073] The data transmission unit (114) transmits location information, area information, and rent information among the structured rental space information to the space analysis unit (121) of the business selection unit (120), and transmits actual measurement dimension information, floor height information, and window location information to the space modeling unit (141) of the model generation unit (140). The data transmission unit (114) optimizes the amount of data transmitted by selectively transmitting only the information required by each unit, and to this end, applies data filtering logic so that unnecessary information is not transmitted to each unit.

[0074] A synchronous data transmission protocol is applied between the data transmission unit (114) and the vendor selection unit (120), so the data transmission unit (114) waits until it receives a confirmation from the vendor selection unit (120) that it has successfully received the data, and if the confirmation is not received within a certain time, it attempts to retransmit. An asynchronous data transmission protocol is applied between the data transmission unit (114) and the model creation unit (140), so that the next task is performed immediately after data transmission, and the model creation unit (140) independently performs the task of creating a 3D model after receiving the data.

[0075] The data transmission unit (114) transmits the company selection information entered through the response provision unit (112) to the matching execution unit (124) of the company selection unit (120). The data transmission unit (114) transmits the company selection information including the building owner's identification information, rental space identification information, the identification information of the selected company, and the selection time information, and provides all information necessary for the matching execution unit (124) to generate matching request data in a batch.

[0076] The data transmission unit (114) transmits the interior company identification information and contact information entered through the interior linkage unit (113) to the communication connection unit (131) of the design input unit (130). The data transmission unit (114) transmits the structured information of the rental space along with the interior company identification information to the design input unit (130), so that when the design input unit (130) performs a communication connection with the representative terminal (P3) of the interior company, the basic information of the rental space can be provided together. Through this, the interior company can identify the structure and characteristics of the rental space in advance before starting the design work, and the accuracy of the design is improved.

[0077] The data transmission unit (114) transmits data to each unit and then generates and stores a transmission completion log. If an error occurs during the data transmission process, the data transmission unit (114) attempts retransmission, and if it fails more than a certain number of times, it displays an error message on the building owner's terminal (P1) and sends a notification to the system administrator. The data transmission unit (114) records the transmission time, receiving unit information, size of the transmitted data, and transmission time in the transmission completion log so that it can be used for system performance monitoring and fault analysis.

[0078] The business selection unit (120) selects a target business based on the rental space information transmitted from the matching unit (110), determines the priority of entry, and performs matching processing. The business selection unit (120) analyzes the location conditions of the rental space, extracts businesses that can enter, evaluates the business viability of each business, and recommends the optimal business. The business selection unit (120) analyzes the rental space information received from the matching unit (110) from various angles to generate objective and quantified evaluation indicators, and through this, maximizes the utilization of the rental space by prioritizing the recommendation of businesses with a high probability of business success to the building owner.

[0079] The vendor selection unit (120) includes a spatial analysis unit (121), a vendor inquiry unit (122), a priority determination unit (123), a matching execution unit (124), and a vendor management unit (125), and these components sequentially process data while forming a mutual feedback loop to derive an optimal matching result.

[0080] The spatial analysis unit (121) receives structured rental space information transmitted from the data transmission unit (114). The spatial analysis unit (121) parses the location information, area information, number of floors information, and rental fee information included in the received rental space information and separates them into respective data items, and a RESTful API communication protocol is used in the connection with the data transmission unit (114) to enable standardized data exchange.

[0081] The spatial analysis unit (121) performs API communication with an external commercial area analysis system (not shown) based on location information included in the received rental space information. The spatial analysis unit (121) transmits the latitude and longitude coordinates of the rental space to the external commercial area analysis system (not shown), and checks the number of pedestrians at the location, the number of surrounding competitors, and the distance to public transportation stations. At this time, the HTTPS protocol is used for communication with the external commercial area analysis system (not shown) to ensure the security of data transmission.

[0082] The External Commercial Area Analysis System (Midosi) provides real-time commercial area data by linking with the Public Data Portal, commercial area information providers, and Geographic Information Systems. The External Commercial Area Analysis System (Midosi) calculates and provides the number of floating populations as the daily average, and calculates and provides the number of nearby competitors as the number of businesses of the same industry located within a radius of 500 meters. The External Commercial Area Analysis System (Midosi) calculates and provides the distance to public transportation stations as the straight-line distance to the nearest subway station or bus stop.

[0083] The spatial analysis unit (121) assigns points based on the number of circulating people. The spatial analysis unit (121) assigns 20 points when the number of circulating people is 10,000 or more, 10 points when it is 5,000 or more but less than 10,000, and 0 points when it is less than 5,000. These scoring criteria are based on statistical evidence showing that, as a result of analyzing commercial space entry data collected over the past three years, the success rate of entry in areas with a circulating population of 10,000 or more was 75% or higher, whereas in areas with a circulating population of less than 5,000, the success rate of entry was 45% or lower. The spatial analysis unit (121) assigns higher scores as the number of circulating people increases, thereby quantitatively reflecting that locations with high circulating populations are advantageous for commercial activities.

[0084] The spatial analysis unit (121) assigns a score based on the number of surrounding competitors found. The spatial analysis unit (121) assigns 20 points when the number of surrounding competitors is less than 3, 10 points when there are 3 or more but less than 5, and 0 points when there are 5 or more. This reflects the market analysis theory that the fewer competitors in the same industry there are, the lower the market saturation, making it easier to acquire customers when opening a new store. The spatial analysis unit (121) assigns a higher score when there are fewer competitors, reflecting that a less competitive location is advantageous for business success, thereby enabling avoidance of excessive competition in a red ocean market.

[0085] The spatial analysis unit (121) assigns a score based on the distance from the public transportation station found. The spatial analysis unit (121) assigns 20 points if the distance from the public transportation station is within 300 meters, 10 points if it is between 300 meters and 500 meters, and 0 points if it is between 500 meters. This reflects the results of a study in which the optimal distance was determined to be within 300 meters as a result of analyzing the impact of pedestrian convenience on the accessibility of commercial facilities in urban planning research. The spatial analysis unit (121) assigns a higher score as public transportation accessibility increases, reflecting that a location with convenient transportation is advantageous for attracting customers.

[0086] The spatial analysis unit (121) sums the scores assigned for the number of floating populations, the number of surrounding competitors, and the distance from public transportation stations. The spatial analysis unit (121) calculates a location score by dividing the summed score by 60 to normalize it to a range of 0 to 100, and the reason for dividing by 60 is that the maximum sum of the scores for the three evaluation items is 60 points (20 points + 20 points + 20 points). By normalizing the location score to a range of 0 to 100, the spatial analysis unit (121) enables the location conditions of different rental spaces to be compared on the same scale, which facilitates the application of weights between each indicator when the priority determination unit (123) evaluates the business feasibility score and the success rate of entry together.

[0087] The spatial analysis unit (121) generates characteristic data for each rental space by combining the calculated location score with the area information and rent information included in the rental space information. The spatial analysis unit (121) includes rental space identification information, location information, area information, rent information, location score, number of floating populations, number of surrounding competitors, and distance from public transportation stations in the characteristic data for each rental space.

[0088] The spatial analysis unit (121) transmits the generated characteristic data for each rental space to the business lookup unit (122) and the priority determination unit (123). In the connection between the spatial analysis unit (121) and the business lookup unit (122), a data sharing memory method is used to store the characteristic data for each rental space in memory, and the business lookup unit (122) refers to the corresponding memory address to minimize data transmission overhead. In the connection between the spatial analysis unit (121) and the priority determination unit (123), an event-based communication method is used so that the spatial analysis unit (121) notifies the priority determination unit (123) as soon as it completes the calculation of the location score, thereby enabling real-time processing. The spatial analysis unit (121) stores the characteristic data for each rental space in a database so that the stored data can be reused when a matching request for the same rental space occurs later, thereby reducing the frequency of API calls with an external commercial area analysis system (not shown) and improving the system response speed.

[0089] The company lookup unit (122) builds a store database (DB2) that stores industry information, average sales information, investment cost information, preferred store location information, preferred area range information, past store success history, and past store failure history.

[0090] The tenant database (DB2) stores each business's unique identification number, business name, brand name, and industry classification code. As an industry classification code, the tenant database (DB2) distinguishes business types including restaurants, cafes, unmanned stores, convenience stores, gyms, academies, clothing stores, and cosmetics stores.

[0091] The vendor database (DB2) stores average sales information for each vendor, including monthly average sales, annual average sales, and average transaction value. The vendor database (DB2) stores investment cost information for each vendor, including initial investment costs, interior costs, equipment investment costs, and initial inventory costs.

[0092] The tenant database (DB2) stores preferred regions, preferred commercial area types, and areas to avoid as information on each business's preferred location. As preferred commercial area types, the tenant database (DB2) stores residential areas, commercial areas, business areas, university districts, and entertainment districts.

[0093] The tenant database (DB2) stores the minimum and maximum preferred areas as information on the preferred area range of each business. The tenant database (DB2) stores the location, area, opening date, and sales performance of successful stores as a history of past successful openings, and stores the location, area, opening date, closing date, and reason for closure of stores that closed down after opening as a history of past failed openings.

[0094] The business lookup unit (122) looks up information for each business stored in the established store database (DB2). The business lookup unit (122) sequentially looks up information for all businesses stored in the store database (DB2) and extracts business type information, preferred store location information, and preferred area range information for each business.

[0095] The business lookup unit (122) extracts potential tenants by comparing them with characteristic data for each rental space transmitted from the space analysis unit (121). The business lookup unit (122) determines whether the location of the rental space corresponds to the tenant's preferred location and whether the area of ​​the rental space falls within the tenant's preferred area range, and classifies only those tenants that satisfy both of these conditions as potential tenants. In the connection between the business lookup unit (122) and the space analysis unit (121), the business lookup unit (122) refers to the characteristic data for each rental space generated by the space analysis unit (121) in real time, and through this, the business lookup unit (122) can filter tenants based on the latest location analysis results.

[0096] The company lookup unit (122) compares the location information of the rental space with the company's preferred area information and classifies the rental space as a company eligible for entry if it belongs to the area preferred by the company. The company lookup unit (122) classifies the company as an eligible company if the area of ​​the rental space is greater than or equal to the company's minimum preferred area and less than or equal to the maximum preferred area.

[0097] The company lookup unit (122) generates store entry condition data for each extracted eligible company, including industry information, average sales information, investment cost information, number of past store entry attempts, and number of past store entry successes. The company lookup unit (122) calculates the number of past store entry attempts by summing the store entry success history and store entry failure history, and calculates the number of past store entry successes as the number of store entry success historys.

[0098] The company lookup unit (122) transmits the generated store entry condition data for each company to the priority determination unit (123). The company lookup unit (122) transmits the store entry condition data for each company by including the unique identification number, company name, and brand name of each company, so that the priority determination unit (123) can clearly identify the companies. A pipeline processing method is used in the connection between the company lookup unit (122) and the priority determination unit (123), so that the company lookup unit (122) transmits the store entry condition data for each company to the priority determination unit (123) as soon as it is generated, and the priority determination unit (123) processes the received data sequentially to determine the priority.

[0099] The priority determination unit (123) combines the characteristic data for each rental space transmitted from the space analysis unit (121) and the entry condition data for each company transmitted from the company inquiry unit (122). For each potential company, the priority determination unit (123) evaluates the business feasibility when the company enters the rental space, and at this time, comprehensively considers the location score calculated by the space analysis unit (121), the characteristics for each company provided by the company inquiry unit (122), and the success rate of entry calculated by the company management unit (125).

[0100] The priority determination unit (123) calculates a business feasibility score for each company using a pre-trained first artificial intelligence model (AI1). The first artificial intelligence model (AI1) is an artificial intelligence model that has learned past store entry success data, average sales data by industry, commercial area analysis data by region, and profitability data relative to rent.

[0101] The first artificial intelligence model (AI1) has a deep neural network structure and consists of an input layer, four hidden layers, and an output layer. The input layer of the first artificial intelligence model (AI1) contains 784 neurons and receives input information regarding the location (latitude, longitude), area, rent, location score, pedestrian traffic, number of nearby competitors, distance from public transportation stations, business type information, average sales information, and investment cost information of the rental space. Each hidden layer of the first artificial intelligence model (AI1) contains 512, 256, 128, and 64 neurons, respectively, and uses the ReLU (Rectified Linear Unit) function as the activation function. The output layer of the first artificial intelligence model (AI1) contains three neurons and outputs expected monthly sales, expected payback period, and expected operating profit margin.

[0102] The training data for the first artificial intelligence model (AI1) consists of 15,000 store entry cases collected over the past five years, of which 10,000 are successful store entry cases (operation continued for more than 6 months) and 5,000 are failed store entry cases (closed within 6 months). For each case, the GPS coordinates of the store location, the leased space area, the monthly rent, the industry code, the actual average monthly sales, the actual payback period, and the actual operating profit margin are assigned as labels.

[0103] The first artificial intelligence model (AI1) is trained using a supervised learning method and uses the Mean Squared Error (MSE) as the loss function during the training process. The loss function of the first artificial intelligence model (AI1) is defined as follows:

[0104] MSE = (1 / n) × Σ(y_actual - y_predicted)²

[0105] Here, n is the number of training data, y_actual is the actual value, and y_predicted is the value predicted by the model.

[0106] The Adam (Adaptive Moment Estimation) optimizer is used for training the first AI model (AI1), and the learning rate is set to 0.001. The first AI model (AI1) is trained for 100 epochs, and Mini-batch Gradient Descent is applied with a batch size of 32 for each epoch. The first AI model (AI1) applies the Dropout technique to prevent overfitting during training, and sets a dropout rate of 0.3 for each hidden layer.

[0107] After the training of the first AI model (AI1) is completed, the performance of the model is evaluated using a validation dataset (3,000 store entry cases not used in training). The accuracy of the first AI model (AI1) in predicting monthly sales was measured at 8.5% based on Mean Absolute Error (MAE), the accuracy of predicting the expected payback period was measured at 12.3% based on MAE, and the accuracy of predicting the expected operating profit margin was measured at 9.7% based on MAE. These prediction accuracy levels are sufficient for business feasibility assessment, and the first AI model (AI1) continuously improves its prediction accuracy by being retrained regularly (every 3 months) by adding new store entry case data.

[0108] The first AI model (AI1) learns from data on past actual store openings to predict the business viability of combinations of specific leased spaces and specific business sectors. The first AI model (AI1) learns by matching input data with actual business performance data, thereby quantitatively learning the impact of the leased space's location conditions and business characteristics on business performance.

[0109] The priority determination unit (123) calculates the expected monthly sales, expected investment recovery period, and expected operating profit margin for each company through the first artificial intelligence model (AI1). The priority determination unit (123) calculates a business feasibility score by assigning a higher value as the expected monthly sales are higher, the expected investment recovery period is shorter, and the expected operating profit margin is higher, thereby ensuring that companies with excellent profitability relative to investment receive a high business feasibility score.

[0110] The priority determination unit (123) calculates the ratio of the expected monthly sales to the standard monthly sales, the ratio of the standard investment recovery period to the expected investment recovery period, and the ratio of the expected operating profit margin to the standard operating profit margin, respectively, and calculates a business feasibility score by weighting the average of these. The priority determination unit (123) normalizes the calculated business feasibility score to a range of 0 to 100, thereby enabling comparison with the location score and the store entry success rate on the same scale.

[0111] The priority determination unit (123) calculates the priority weight for store entry based on the calculated business feasibility score, the location score transmitted from the space analysis unit (121), the number of past store entry attempts transmitted from the business lookup unit (122), and the number of past store entry successes.

[0112] Before calculating the priority weight for store entry, the priority determination unit (123) receives the store entry success rate from the business evaluation unit (125b) of the business management unit (125). The priority determination unit (123) uses the received store entry success rate along with the business feasibility score and location score to calculate the priority weight for store entry. At this time, a real-time data synchronization mechanism operates between the priority determination unit (123) and the business evaluation unit (125b), so that when the business evaluation unit (125b) updates the store entry success rate, it is immediately reflected in the priority determination unit (123).

[0113] The priority determination unit (123) calculates the priority weight for store entry according to the following [Equation 1].

[0114] [Mathematical Formula 1]

[0115] W = α × B + β × L + γ × S

[0116] Here, W is the priority weight for store entry, B is the business feasibility score calculated by the priority determination unit (123) (normalized to a value between 0 and 100), L is the location score calculated by the spatial analysis unit (121) (normalized to a value between 0 and 100), and S is the store entry success rate (percentage, between 0 and 100) calculated by the business evaluation unit (125b). α, β, and γ represent weighting coefficients for the business feasibility score, location score, and store entry success rate, respectively, and α, β, and γ are each values ​​between 0 and 1, and α + β + γ = 1. A higher W value indicates a higher priority for store entry.

[0117] By calculating the priority weights for store entry using [Mathematical Formula 1], evaluation indicators of different dimensions—business feasibility score, location score, and store entry success rate—can be considered in an integrated manner. The business feasibility score represents the expected profitability of the business when it enters the leased space, the location score represents the commercial conditions of the leased space itself, and the store entry success rate represents the business's past entry performance. By applying a weighted average to these three indicators, bias that may arise from evaluating based on a single indicator is prevented, and businesses with a high probability of actual business success can be placed at the top through a multifaceted evaluation.

[0118] Experimental results showed that among the top 10 companies selected by applying [Equation 1], 8 companies continued to operate successfully for more than 6 months after opening, whereas only 6 companies were selected based solely on business feasibility scores and only 5 companies were selected based solely on location scores. This means that the integrated evaluation method of [Equation 1] shows a matching success rate approximately 30% higher than the single-indicator evaluation method.

[0119] The priority determination unit (123) may apply weighting coefficients α, β, and γ as values ​​set by the manager, or automatically calculate the optimal weighting coefficients by analyzing past matching success data. In one embodiment, the priority determination unit (123) uses α = 0.5, β = 0.3, and γ = 0.2 as default values, which means that the business feasibility score is considered most important, while the location score and store entry success rate are considered secondarily. These weighting coefficient settings are based on statistical evidence that the business feasibility score has the greatest influence on matching success as a result of analyzing past matching data.

[0120] The priority determination unit (123) sorts the list of eligible businesses according to the entry priority weight calculated for each eligible business. The priority determination unit (123) sorts by placing businesses with high entry priority weights at the top and businesses with low entry priority weights at the bottom, and performs efficient sorting using Quick Sort or Merge Sort as the sorting algorithm.

[0121] The priority determination unit (123) excludes businesses whose calculated entry priority weight is less than a preset minimum priority threshold from the list of eligible businesses. The priority determination unit (123) applies the minimum priority threshold as a value set by the administrator, and in one embodiment, the minimum priority threshold is set to 50. The priority determination unit (123) determines that businesses with an entry priority weight of less than 50 have a low probability of business success and excludes them from the recommendation list, thereby providing the building owner with only a highly reliable recommendation list.

[0122] The priority determination unit (123) transmits the sorted list of prospective tenants to the matching execution unit (124). The priority determination unit (123) transmits the list of prospective tenants including each tenant's priority weight, business feasibility score, location score, and success rate for entry, enabling the matching execution unit (124) to utilize this information when generating matching request data. In the connection between the priority determination unit (123) and the matching execution unit (124), a data serialization method is used to efficiently transmit the list of prospective tenants with a complex structure.

[0123] The matching execution unit (124) transmits the list of prospective tenants sorted by the priority determination unit (123) to the response providing unit (112). The matching execution unit (124) converts the list of prospective tenants into JSON or XML format and transmits it to the response providing unit (112) via a network, and the response providing unit (112) visually displays it on the building owner's terminal (P1). In the connection between the matching execution unit (124) and the response providing unit (112), a RESTful API method is used to enable standardized data exchange, thereby improving system scalability and maintainability.

[0124] The matching execution unit (124) receives the vendor selection information transmitted from the data transmission unit (114). The matching execution unit (124) extracts the identification information of the selected vendor included in the vendor selection information and searches for the head office system contact information of the vendor in the vendor database (DB2).

[0125] The matching execution unit (124) transmits matching request data, including rental space information, building owner contact information, and matching request time information, to the company's headquarters system (not shown). The matching execution unit (124) transmits the matching request data via at least one of email, SMS, or API call, and selects the transmission method according to the company's preferred contact method. In the connection between the matching execution unit (124) and the headquarters system, a secure communication protocol (HTTPS, TLS) is used so that the matching request data is encrypted and transmitted, thereby protecting the personal information of the building owner and the company.

[0126] When the matching execution unit (124) receives a matching acceptance response from the headquarters system, it determines that the matching is complete. After the matching is completed, the matching execution unit (124) generates matching result data including rental space identification information, selected company identification information, matching completion time information, and matching request time information.

[0127] The matching execution unit (124) transmits the generated matching result data to the matching data storage unit (161) of the data management unit (160). The matching execution unit (124) sends a matching completion notification message to the building owner's terminal (P1) and the headquarters system of the selected company to notify that the matching has been successfully completed, and this notification message includes guidance information such as the interior design schedule to proceed to the next step. A transaction management mechanism is applied to the connection between the matching execution unit (124) and the data management unit (160) to ensure data consistency by preventing the matching completion notification from being sent when the matching result data is not properly stored in the data management unit (160).

[0128] The company management unit (125) includes a company data storage unit (125a) and a company evaluation unit (125b).

[0129] The business data storage unit (125a) manages the store database (DB2). When new business information is registered, the business data storage unit (125a) adds the information to the store database (DB2), and when existing business information is changed, it modifies the information in the store database (DB2).

[0130] When the entry history of each company is added, the company data storage unit (125a) updates the number of past attempts to enter and the number of past successful entries. When the company data storage unit (125a) receives matching completion information from the matching execution unit (124), it increases the number of past attempts to enter for the company by 1. When the business continues for a certain period (e.g., 6 months) or longer after entry, the company data storage unit (125a) increases the number of past successful entries for the company by 1, and at this time, the company data storage unit (125a) periodically receives information on whether the business of the entered store continues from the data management unit (160) to determine whether the entry is successful.

[0131] The company evaluation unit (125b) calculates the entry success rate based on the number of past attempts to enter and the number of past successful entries for each company transmitted from the company inquiry unit (122). If the number of past attempts to enter is 0, the company evaluation unit (125b) assigns a basic entry success rate of 50. The company evaluation unit (125b) assigns a basic value to new companies by assuming an average success rate, thereby ensuring that new companies are not evaluated too unfavorably, and through this, fair entry opportunities are given to new companies as well.

[0132] If the number of past attempts to enter a store is 1 or more, the company evaluation (125b) calculates the store entry success rate as a percentage by multiplying the value obtained by dividing the number of past successful store entries by the number of past attempts to enter a store by 100. For example, if the number of past attempts to enter a store is 10 and the number of past successful store entries is 7, the company evaluation (125b) calculates the store entry success rate as (7 / 10) × 100 = 70.

[0133] The business evaluation unit (125b) transmits the calculated store entry success rate to the priority determination unit (123). The business evaluation unit (125b) transmits the store entry success rate along with the identification information of each business so that the priority determination unit (123) can utilize it when calculating the store entry priority weight of the business. A caching mechanism is applied in the connection between the business evaluation unit (125b) and the priority determination unit (123) to improve the search speed by caching the store entry success rate information of frequently searched businesses in memory.

[0134] 3. Industry Database and Calculation of Business Feasibility Score

[0135] One embodiment of the present invention further includes an industry database (DB3) in which the average conversion rate, average transaction value, average cost ratio, standard monthly sales, standard investment recovery period, and standard monthly net profit are stored for each industry, including restaurants, cafes, unmanned stores, convenience stores, gyms, and academies.

[0136] The industry database (DB3) stores the average conversion rate for each industry. The average conversion rate refers to the proportion of customers who actually enter a store of the corresponding industry and make a purchase among the floating population. For example, the industry database (DB3) stores the average conversion rate for restaurants as 3%, the average conversion rate for cafes as 5%, and the average conversion rate for convenience stores as 10%.

[0137] The industry database (DB3) stores the average transaction value for each industry. The average transaction value refers to the average payment amount per customer. For example, the industry database (DB3) stores the average transaction value for restaurants as 15,000 won, the average transaction value for cafes as 6,000 won, and the average transaction value for convenience stores as 3,000 won.

[0138] The industry database (DB3) stores the average cost ratio for each industry. The average cost ratio refers to the ratio of operating costs to sales and includes labor costs, material costs, administrative costs, and marketing costs. For example, the industry database (DB3) stores the average cost ratio for restaurants as 70%, for cafes as 60%, and for convenience stores as 80%.

[0139] The industry database (DB3) stores the standard monthly sales for each industry. Standard monthly sales refer to the average monthly sales achieved in the corresponding industry. For example, the industry database (DB3) stores the standard monthly sales for restaurants as 30,000,000 won, the standard monthly sales for cafes as 20,000,000 won, and the standard monthly sales for convenience stores as 25,000,000 won.

[0140] The industry database (DB3) stores the standard payback period for each industry. The standard payback period refers to the average time it takes to recover the initial investment costs in the corresponding industry and is displayed in months. For example, the industry database (DB3) stores the standard payback period for restaurants as 24 months, for cafes as 18 months, and for convenience stores as 30 months.

[0141] The industry database (DB3) stores the standard monthly net profit for each industry. The standard monthly net profit refers to the average net profit remaining after deducting monthly operating expenses and monthly rent from monthly sales. For example, the industry database (DB3) stores the standard monthly net profit for restaurants as 5,000,000 won, the standard monthly net profit for cafes as 4,000,000 won, and the standard monthly net profit for convenience stores as 3,000,000 won.

[0142] The priority determination unit (123) further includes a commercial area data storage unit (123a), a sales forecasting unit (123b), a profitability analysis unit (123c), and a score calculation unit (123d).

[0143] The commercial area data storage unit (123a) stores information regarding the number of floating populations, the number of surrounding competitors, and the distance from public transportation stations, which are included in the characteristic data for each rental space transmitted from the spatial analysis unit (121). The commercial area data storage unit (123a) stores this information along with rental space identification information so that it can be quickly retrieved during subsequent business feasibility analysis. A real-time data synchronization mechanism operates between the commercial area data storage unit (123a) and the spatial analysis unit (121), so that when the spatial analysis unit (121) updates the commercial area data, it is immediately reflected in the commercial area data storage unit (123a).

[0144] The commercial area data storage unit (123a) periodically updates the stored commercial area data. The commercial area data storage unit (123a) communicates periodically (weekly or monthly) with an external commercial area analysis system (not shown) to update information on the latest floating population, the number of surrounding competitors, and the distance to public transportation stations. The commercial area data storage unit (123a) records changes by comparing the updated information with existing information, and if the changes exceed a certain level (e.g., a change in floating population of 20% or more), it sends a notification to the system administrator to notify that a re-evaluation of the location conditions of the leased space is necessary.

[0145] The sales prediction unit (123b) retrieves the average conversion rate and average transaction value of the corresponding industry from the industry database (DB3) based on the industry information included in the store entry condition data for each company transmitted from the company inquiry unit (122). The sales prediction unit (123b) calculates the expected sales using the retrieved average conversion rate and average transaction value, and a database indexing technique is applied to the connection between the sales prediction unit (123b) and the industry database (DB3) so that the average conversion rate and average transaction value can be retrieved quickly based on the industry information.

[0146] The sales prediction unit (123b) calculates the expected number of customers by multiplying the number of people in the commercial area data storage unit (123a) by the average conversion rate retrieved. For example, if the number of people in the commercial area data storage unit (123b) is 10,000 and the average conversion rate is 5%, the sales prediction unit (123b) calculates the expected number of customers as 10,000 × 0.05 = 500.

[0147] The sales forecasting unit (123b) calculates the expected daily sales by multiplying the estimated number of customers by the average transaction value. For example, if the estimated number of customers is 500 and the average transaction value is 6,000 won, the sales forecasting unit (123b) calculates the expected daily sales as 500 × 6,000 = 3,000,000 won.

[0148] The sales forecasting unit (123b) calculates the expected monthly sales by multiplying the expected daily sales by the number of business days. The sales forecasting unit (123b) may assume the number of business days per month is 30 days or apply the average number of business days per industry. For example, if the expected daily sales are 3,000,000 won and the number of business days per month is 30 days, the sales forecasting unit (123b) calculates the expected monthly sales as 3,000,000 × 30 = 90,000,000 won.

[0149] The sales forecasting unit (123b) transmits the calculated estimated monthly sales to the profitability analysis unit (123c) and the score calculation unit (123d). The sales forecasting unit (123b) transmits the estimated number of customers, average transaction value, number of floating populations, and average conversion rate along with the estimated monthly sales, thereby ensuring transparency in the profitability analysis, and through this, the profitability analysis unit (123c) can track each step of the sales calculation process. A data pipeline method is used in the connection between the sales forecasting unit (123b) and the profitability analysis unit (123c), so that as soon as the sales forecasting unit (123b) calculates the estimated monthly sales, it is transmitted to the profitability analysis unit (123c), enabling real-time processing.

[0150] The profitability analysis unit (123c) analyzes profitability based on investment cost information included in the store entry condition data for each company transmitted from the company inquiry unit (122) and the expected monthly sales calculated from the sales prediction unit (123b).

[0151] The profitability analysis unit (123c) calculates monthly operating costs by multiplying the expected monthly sales by the average cost ratio by industry retrieved from the industry database (DB3). The profitability analysis unit (123c) assumes that the monthly operating costs include labor costs, material costs, administrative costs, and marketing costs. For example, if the expected monthly sales are 90,000,000 won and the average cost ratio is 60%, the profitability analysis unit (123c) calculates the monthly operating costs as 90,000,000 × 0.6 = 54,000,000 won.

[0152] The profitability analysis unit (123c) calculates the monthly net profit by subtracting the monthly rent and calculated monthly operating costs entered through the information input unit (111) from the expected monthly sales. For example, if the expected monthly sales are 90,000,000 won, the monthly rent is 10,000,000 won, and the monthly operating costs are 54,000,000 won, the profitability analysis unit (123c) calculates the monthly net profit as 90,000,000 - 10,000,000 - 54,000,000 = 26,000,000 won, and through this, the net profit that the company can actually obtain when it enters the leased space can be quantitatively evaluated.

[0153] The profitability analysis unit (123c) calculates the payback period by dividing the initial investment cost included in the investment cost information by the monthly net profit. The profitability analysis unit (123c) calculates the payback period in months. For example, if the initial investment cost is 100,000,000 won and the monthly net profit is 5,000,000 won, the profitability analysis unit (123c) calculates the payback period as 100,000,000 / 5,000,000 = 20 months.

[0154] The lack of profitability analysis (123c) determines that if the monthly net profit is negative, the company is expected to incur a deficit and sets the investment recovery period to infinite. The lack of profitability analysis (123c) assigns a very low business feasibility score to companies expected to incur a deficit, thereby excluding them from the list of eligible tenants, and thereby prevents both the building owner and the company from suffering economic losses in advance.

[0155] The profitability analysis unit (123c) transmits the calculated investment recovery period and monthly net profit to the score calculation unit (123d). The profitability analysis unit (123c) transmits the monthly operating costs, monthly rent, and initial investment costs along with the investment recovery period and monthly net profit to clarify the basis for calculating the business feasibility score. In the connection between the profitability analysis unit (123c) and the score calculation unit (123d), a data verification mechanism operates to check whether the values ​​calculated by the profitability analysis unit (123c) are logically valid (e.g., whether the monthly net profit does not exceed the expected monthly sales) and then transmits them to the score calculation unit (123d).

[0156] The score calculation unit (123d) calculates a business feasibility score by combining the expected monthly sales calculated from the sales forecast unit (123b), the investment recovery period calculated from the profitability analysis unit (123c), and the monthly net profit.

[0157] The score calculation member (123d) calculates the business feasibility score by the following [Equation 2].

[0158] [Mathematical Formula 2]

[0159] B = δ × (R / R_ref) × 100 + ε × (P_ref / P) × 100 + ζ × (M / M_ref) × 100

[0160] Here, B is the business feasibility score, R is the expected monthly sales calculated from the sales forecasting component (123b), R_ref is the standard monthly sales of the corresponding industry stored in the industry database (DB3), P is the investment recovery period (in months) calculated from the profitability analysis component (123c), P_ref is the standard investment recovery period (in months) of the corresponding industry stored in the industry database (DB3), M is the monthly net profit calculated from the profitability analysis component (123c), and M_ref is the standard monthly net profit of the corresponding industry stored in the industry database (DB3). δ, ε, and ζ represent weighting coefficients for the expected monthly sales, investment recovery period, and monthly net profit, respectively, and δ, ε, and ζ are values ​​greater than or equal to 0 and less than or equal to 1, respectively, and δ + ε + ζ = 1. The higher the final business feasibility score, which is adjusted from 0 to 100, the higher the business feasibility.

[0161] By calculating the business feasibility score using [Equation 2], different financial indicators such as projected monthly sales, payback period, and monthly net profit can be evaluated in an integrated manner. Projected monthly sales represent the sales volume of the store, the payback period represents investment efficiency, and monthly net profit represents actual profitability. By normalizing each indicator in [Equation 2] by dividing it by the standard value of the corresponding industry, comparisons between different industries become possible. For example, although it is difficult to directly compare the projected monthly sales of 50,000,000 won for a restaurant with the projected monthly sales of 30,000,000 won for a cafe, normalizing them by dividing them by their respective standard monthly sales (30,000,000 won for the restaurant and 20,000,000 won for the cafe) yields a sales ratio of 1.67 for the restaurant and 1.5 for the cafe, allowing for a comparison of relative superiority.

[0162] In particular, calculating the payback period item in the form of P_ref / P is intended to assign higher scores as the payback period is shorter. If the payback period is shorter than the threshold value, the P_ref / P value becomes greater than 1, resulting in a high score; conversely, if the payback period is longer than the threshold value, the P_ref / P value becomes less than 1, resulting in a low score. This allows for awarding bonus points to companies with short payback periods, i.e., companies with high investment efficiency.

[0163] As a result of the experiment, the correlation coefficient between the business feasibility score calculated by applying [Equation 2] and the actual business performance after entry was measured to be 0.78, which means that the business feasibility score of [Equation 2] predicts actual business performance with considerable accuracy. For businesses with a business feasibility score of 80 or higher, the survival rate after one year of entry was 85% or higher, while for businesses with a business feasibility score of less than 50, the survival rate after one year of entry was 40% or lower, verifying that the business feasibility score of [Equation 2] is a valid indicator for judging the likelihood of successful entry.

[0164] The score calculation unit (123d) calculates a business feasibility score by applying [Mathematical Formula 2], and then adjusts the calculated business feasibility score to 100 if it exceeds 100 and to 0 if it is less than 0 to generate a final business feasibility score. By limiting the business feasibility score to a range of 0 or more and 100 or less, the score calculation unit (123d) enables the business feasibility of different companies to be compared using the same standard.

[0165] The score calculation unit (123d) applies weighting coefficients δ, ε, and ζ as values ​​set by the manager. In one embodiment, the score calculation unit (123d) uses δ = 0.4, ε = 0.3, and ζ = 0.3 as default values, which means that the expected monthly sales are considered most important, while the payback period and monthly net profit are considered equally. These weighting coefficient settings are based on statistical evidence that, as a result of analyzing past business performance data, the sales volume has the greatest impact on business sustainability.

[0166] The score calculation unit (123d) transmits the generated final business feasibility score to the priority determination unit (123). Along with the final business feasibility score, the score calculation unit (123d) transmits the expected monthly sales, investment recovery period, and monthly net profit so that the priority determination unit (123) can utilize them when calculating the priority weight for store entry. Through this, the building owner can check the detailed financial indicators of each business and make a final decision. A real-time data streaming method is used in the connection between the score calculation unit (123d) and the priority determination unit (123), so that the score calculation unit (123d) transmits the final business feasibility score to the priority determination unit (123) as soon as it is generated, and the priority determination is performed without delay.

[0167] The priority determination unit (123) calculates the entry priority weight W by using the final business feasibility score transmitted from the score calculation unit (123d) as the input value B of [Equation 1]. The priority determination unit (123) calculates the entry priority weight of each company by substituting the final business feasibility score, location score, and entry success rate into [Equation 1], thereby performing a comprehensive evaluation that considers business feasibility, location conditions, and past performance.

[0168] The priority determination unit (123) excludes companies from the list of eligible tenants whose final business feasibility score transmitted from the score calculation unit (123d) is less than a preset minimum business feasibility standard score. The priority determination unit (123) applies the minimum business feasibility standard score as a value set by the manager, and in one embodiment, the minimum business feasibility standard score is set to 30. The priority determination unit (123) determines that companies with a final business feasibility score of less than 30 have a very low probability of business success and excludes them from the recommendation list, thereby preventing the building owner and the company from wasting unnecessary time and money.

[0169] The design input unit (130) receives interior design drawing data from the representative terminal (P3) of the interior company selected through the matching unit (110). The design input unit (130) establishes a communication connection with the representative terminal (P3) of the interior company, receives 2D interior design drawing data, converts it into a structured design data format, and transmits the converted data to the model generation unit (140) to provide a basis for creating a 3D virtual interior model. The design input unit (130) automatically establishes a communication connection based on the interior company information received from the matching unit (110), thereby enabling the building owner or the interior company to upload the design drawing without separate complex settings.

[0170] The design input unit (130) includes a communication connection unit (131), a design data receiving unit (132), and a data conversion unit (133), and these components sequentially process data and convert a 2D design drawing provided by an interior design company into a format suitable for creating a 3D model.

[0171] The communication connection unit (131) performs a communication connection with the representative terminal (P3) of the interior company based on the interior company identification information and contact information transmitted from the data transmission unit (114). The communication connection unit (131) transmits a connection request message to the representative terminal (P3) of the interior company using the email address or phone number included in the interior company identification information. At this time, a message queue method is used in the connection between the communication connection unit (131) and the data transmission unit (114) so ​​that the communication connection unit (131) can process information sequentially even if the data transmission unit (114) transmits information from multiple interior companies.

[0172] The communication connection unit (131) includes rental space identification information, building owner contact information, and a blueprint upload link in the connection request message. When the blueprint upload link is clicked, the communication connection unit (131) causes a file upload interface to be displayed on the interior design company's representative terminal (P3), and allows the blueprint file to be uploaded through the interface. The communication connection unit (131) includes the rental space identification information as a URL parameter in the blueprint upload link so that the uploaded blueprint can be automatically identified as belonging to which rental space.

[0173] When the communication connection unit (131) receives a request to upload a design drawing from the representative terminal (P3) of the interior design company, it establishes a secure file transfer channel using a secure communication protocol (HTTPS, SFTP). The communication connection unit (131) receives the design drawing file through the file transfer channel, and calculates an MD5 or SHA-256 hash value to verify the integrity of the file during the reception process and compares the hash values ​​before and after transmission. In the connection between the communication connection unit (131) and the representative terminal (P3) of the interior design company, the TLS 1.3 protocol is applied so that encryption is performed during data transmission, thereby protecting the design drawing file from being exposed to the outside during the transmission process.

[0174] The communication connection unit (131) stores the received design drawing file in a temporary storage and then transmits the file path information to the design data receiving unit (132). When the file upload is completed, the communication connection unit (131) sends an upload completion notification message to the representative terminal (P3) of the interior design company, and this notification message includes the name, size, and upload time information of the uploaded file. In the connection between the communication connection unit (131) and the design data receiving unit (132), rental space identification information and interior design company identification information are transmitted along with the file path information so that the design data receiving unit (132) can clearly identify which project the design drawing belongs to.

[0175] The design data receiving unit (132) receives interior design drawing data including at least one of a 2D floor plan, elevation view, and cross-section view from the representative terminal (P3) of the interior company through the communication connection unit (131). The design data receiving unit (132) checks the file format of the received interior design drawing data and supports DWG, DXF, PDF, JPG, and PNG formats. The design data receiving unit (132) checks the file extension and analyzes the file header to verify whether the actual file format matches the extension, thereby preventing malicious or damaged files from being processed.

[0176] The design data receiving unit (132) extracts wall information, furniture placement information, finishing material information, and lighting information from the received interior design drawing data. If the interior design drawing data is in a CAD file format (DWG, DXF), the design data receiving unit (132) classifies objects by layer using a CAD file parsing library (e.g., the Teigha library of the Open Design Alliance or the parsing engine of LibreCAD) and extracts coordinate, size, and attribute information of each object. The design data receiving unit (132) analyzes the layer names of the CAD file and classifies layers containing keywords such as "WALL", "FURNITURE", "MATERIAL", and "LIGHTING" into wall, furniture, finishing material, and lighting layers, respectively.

[0177] The design data receiving unit (132) extracts the starting coordinates, ending coordinates, thickness, height, and material of the wall as wall information. If the wall object is represented in the form of a polyline or a line, the design data receiving unit (132) extracts the starting and ending coordinates of each line segment, and the wall thickness is extracted from the object's line weight attribute or separate attribute data. If the wall height information is not specified in the CAD file, the design data receiving unit (132) uses the floor height information of the rental space received from the data transmission unit (114) as the default value.

[0178] The design data receiving unit (132) extracts the center coordinates, width, length, height, rotation angle, and furniture type of each piece of furniture as furniture placement information. When a furniture object is expressed in the form of a block or an insert, the design data receiving unit (132) uses the insertion point coordinates of the block as the center coordinates of the furniture and uses the size of the block's bounding box as the width and length of the furniture. The design data receiving unit (132) identifies the furniture type (table, chair, sofa, etc.) by analyzing the block name or attribute data, and if the block name contains keywords such as "TABLE", "CHAIR", "SOFA", etc., it extracts the corresponding keywords as the furniture type.

[0179] The design data receiving unit (132) extracts the type, color, and texture image of the finishing material applied to the wall, floor, and ceiling as finishing material information. If the finishing material is expressed in the form of a hatch pattern or a region, the design data receiving unit (132) maps the type of hatch pattern to the type of finishing material; for example, a brick pattern is mapped to a brick finishing material, and a wood grain pattern is mapped to a wood finishing material. The design data receiving unit (132) extracts color information from the object's color attribute and converts the color information stored in RGB or ACI (AutoCAD Color Index) format into a standard RGB format.

[0180] The design data receiving unit (132) extracts the position coordinates of the lighting, the type of lighting, the brightness of the lighting, and the color temperature as lighting information. When the lighting object is expressed in the form of a point or a symbol, the design data receiving unit (132) uses the coordinates of the point as the position coordinates of the lighting and identifies the type of lighting (ceiling light, spotlight, indirect lighting, etc.) by analyzing the symbol name or attribute data. If the lighting brightness and color temperature information is not specified in the CAD file, the design data receiving unit (132) applies standard brightness and color temperature values ​​for each type of lighting.

[0181] When the interior design drawing data is in an image file format (PDF, JPG, PNG), the design data receiving unit (132) uses an image recognition algorithm to recognize walls, furniture, and lighting in the design drawing image and estimates the location and size of each object. The design data receiving unit (132) can use computer vision technology and deep learning-based object detection models (e.g., YOLO, Faster R-CNN, Mask R-CNN) as the image recognition algorithm, and these models are pre-trained with thousands of past interior design drawing images to detect objects such as walls, furniture, and lighting with high accuracy. The design data receiving unit (132) extracts the bounding box coordinates of the detected objects and refers to the scale information displayed on the design drawing image to convert them into actual spatial coordinates.

[0182] The design data receiving unit (132) transmits the extracted wall information, furniture arrangement information, finishing material information, and lighting information to the data conversion unit (133). If there is information that is missing or unclear during the extraction process, the design data receiving unit (132) transmits a message requesting additional information input to the representative terminal (P3) of the interior design company, and this message provides guidance on the missing information items and input methods. In the connection between the design data receiving unit (132) and the data conversion unit (133), the extracted information is serialized and transmitted in JSON or XML format, thereby enabling the data conversion unit (133) to receive data in a standardized format.

[0183] The data conversion unit (133) converts wall information, furniture arrangement information, finishing material information, and lighting information extracted from the design data receiving unit (132) into a structured design data format. The data conversion unit (133) converts the extracted information into structured data in JSON or XML format and assigns a unique identification number to each object so that the model generation unit (140) can clearly identify each object.

[0184] When the data conversion unit (133) converts wall information into a structured format, it represents the starting coordinates of each wall as (x1, y1, z1) and the ending coordinates as (x2, y2, z2), and adds thickness, height, and material information as attributes. If the coordinates extracted from the 2D design include only x and y coordinates, the data conversion unit (133) converts them into a 3D coordinate format by setting the z coordinate to 0, and the height of the wall is represented as a vector in the z direction. The data conversion unit (133) assigns a unique identification number to each wall and adds the wall type (exterior wall, interior wall, partition, etc.) as an attribute so that the internal modeling unit (142) can properly process the wall.

[0185] When the data conversion unit (133) converts furniture arrangement information into a structured format, it expresses the center coordinates of each piece of furniture as (x, y, z) and adds the width, length, height, rotation angle, and furniture type as attributes. The data conversion unit (133) normalizes the rotation angle of the furniture to a value between 0 and 360 degrees, and defines the rotation angle as a rotation based on the z-axis. The data conversion unit (133) maps the furniture type to standard furniture categories (table, chair, sofa, bed, cabinet, countertop, refrigerator, air conditioner, etc.) and classifies furniture that does not belong to a standard category into an "Other" category.

[0186] When converting finishing material information into a structured format, the data conversion unit (133) adds the type of finishing material, color code (RGB or HEX), and texture image file path as attributes, along with the coordinate range of the application target area (wall, floor, ceiling). The data conversion unit (133) represents the application target area in the form of a polygon and stores the coordinates of each vertex of the polygon in an array. The data conversion unit (133) unifies the color code into the RGB format and converts color information stored in ACI or other color formats into the RGB format. If a texture image is provided, the data conversion unit (133) stores the image file in the system's internal storage and includes the storage path in the structured data.

[0187] When converting lighting information into a structured format, the data conversion unit (133) expresses the position coordinates of the lighting as (x, y, z) and adds the lighting type, lighting brightness (in lumens), and color temperature (in Kelvin) as attributes. The data conversion unit (133) maps the lighting type to standard lighting types (point light, spotlight, directional light, area light, etc.) and sets additional attributes according to each lighting type (illumination angle in the case of a spotlight, area in the case of an area light, etc.). The data conversion unit (133) standardizes the lighting brightness to lumens and, if provided in watts, applies a conversion factor according to the lighting type to convert it to lumens.

[0188] The data conversion unit (133) transmits the converted structured design data to the space modeling unit (141), internal modeling unit (142), external modeling unit (143), and rendering unit (144) of the model generation unit (140). The data conversion unit (133) transmits the rental space identification information, interior design company identification information, and design drawing reception time information together with the structured design data so that the model generation unit (140) can properly process the design data. At this time, a message broker method is used in the connection between the data conversion unit (133) and the model generation unit (140) so that the data conversion unit (133) transmits data asynchronously and the model generation unit (140) can process the data independently.

[0189] If an error occurs during the conversion process, the data conversion unit (133) records the error details in a log and sends a re-extraction request to the design data receiving unit (132). If the re-extraction fails more than a certain number of times, the data conversion unit (133) sends a request message for re-uploading the design drawing to the representative terminal (P3) of the interior company, and this message includes the error details along with guidance on the correct design drawing format. The data conversion unit (133) sends a signal indicating that the design data has been received from the design input unit (130) to the progress status management unit (164), so that the progress status management unit (164) can update the interior progress stage to the "design progress stage."

[0190] The model generation unit (140) inputs the interior design drawing data transmitted from the design input unit (130) into a pre-trained artificial intelligence model to generate a three-dimensional virtual interior model including the interior and exterior of the rental space. The model generation unit (140) models the basic structure of the rental space in three dimensions, arranges walls and furniture, generates an exterior model, and applies textures and lighting to complete the final three-dimensional virtual interior model. The model generation unit (140) generates an accurate three-dimensional model by combining the structured design data received from the design input unit (130) and the actual measurement information of the rental space received from the matching unit (110), thereby enabling the building owner and the customer to visually check the actual construction results in advance.

[0191] The model generation unit (140) includes a spatial modeling unit (141), an internal modeling unit (142), an external modeling unit (143), a rendering unit (144), and an AI processing unit (145), and these components sequentially process data to convert a two-dimensional design drawing into a high-quality three-dimensional virtual model.

[0192] The space modeling unit (141) receives structured design data transmitted from the data conversion unit (133) and receives actual measurement dimension information, floor height information, and window location information of the rental space transmitted from the data transmission unit (114). The space modeling unit (141) parses the received information to extract the width, length, and height of the rental space, and a real-time data streaming method is used in connection with the data conversion unit (133) so that the space modeling unit (141) receives and begins processing the structured design data as soon as the data conversion unit (133) transmits it.

[0193] The space modeling unit (141) generates a basic space model in the shape of a three-dimensional rectangular prism having width, length, and height based on received actual measurement information and floor height information. The space modeling unit (141) uses a three-dimensional graphics library (e.g., OpenGL, DirectX, Three.js, Blender Python API) to generate six faces (floor, ceiling, and four walls) of the rectangular prism as polygon meshes, and each face is composed of a set of triangular or quadrilateral polygons. The space modeling unit (141) places the floor face on the z=0 plane, places the ceiling face on the z=floor height plane, and places the four walls parallel to the x-axis and y-axis to form the shape of a rectangular prism.

[0194] The space modeling unit (141) completes a three-dimensional space basic model by forming an opening at a corresponding location according to the received window location information. The space modeling unit (141) creates a window-shaped opening on the corresponding wall surface using the wall surface identification information (east wall, west wall, south wall, north wall), the center coordinates of the window, the horizontal length, and the vertical length of the window included in the window location information. When creating the opening, the space modeling unit (141) represents the window by deleting the polygon mesh of the corresponding area through a Boolean operation or by applying a transparent material to the corresponding area. The space modeling unit (141) tessellates the polygon mesh around the window opening so that the boundary of the opening is smoothly represented.

[0195] The space modeling unit (141) applies a glass material to the window opening to set transparency, reflectance, and refractive index. The space modeling unit (141) sets the transparency of the glass material to a range of 7090% to allow external light rays to pass through to the interior, sets the reflectance to a range of 1020% to represent reflection on the glass surface, and sets the refractive index to 1.5 to simulate the refraction of light rays passing through the glass. The space modeling unit (141) creates a window frame and places it around the opening to realize a shape similar to an actual window, and uses information or standard values ​​included in the design data for the thickness, color, and material of the window frame.

[0196] The spatial modeling unit (141) transmits the completed 3D spatial basic model to the internal modeling unit (142) and the external modeling unit (143). The spatial modeling unit (141) transmits the 3D spatial basic model data by storing it in one of the FBX, OBJ, or GLTF formats, and these formats are widely used as standard formats for data exchange between 3D graphic software. In the connection between the spatial modeling unit (141) and the internal modeling unit (142), a shared memory method is used to store the 3D spatial basic model data in memory, and the internal modeling unit (142) references the corresponding memory address to minimize data copying overhead. In the connection between the spatial modeling unit (141) and the external modeling unit (143), the same 3D spatial basic model data is shared, and the external modeling unit (143) selectively processes only the outer wall area.

[0197] The internal modeling unit (142) receives a three-dimensional space basic model generated from the space modeling unit (141). The internal modeling unit (142) extracts wall information included in the structured design data transmitted from the data conversion unit (133) and arranges walls that partition the internal space.

[0198] The internal modeling unit (142) creates an internal wall model using the wall's start coordinates, end coordinates, thickness, and height included in the wall information. The internal modeling unit (142) creates each wall as a rectangular polygon mesh and arranges the walls in a direction connecting the start coordinates and the end coordinates. The internal modeling unit (142) sets the width of the polygon mesh to the thickness of the wall, sets the height of the polygon mesh to the height of the wall, and sets the z-direction range so that the wall starts from the floor and extends to the ceiling. The internal modeling unit (142) appropriately merges the polygon meshes at the points where the walls intersect to ensure that the connection between the walls is natural.

[0199] The interior modeling unit (142) creates an interior space partition model by placing the generated interior wall model onto a three-dimensional space basic model. The interior modeling unit (142) merges polygon meshes at the parts where the wall contacts the floor and ceiling so that the wall and the space structure are naturally connected, and uses Boolean operations to remove overlapping areas. If there is a door opening, the interior modeling unit (142) creates an opening at the corresponding location on the wall and places a door frame and door leaf model.

[0200] The internal modeling unit (142) places furniture objects in three-dimensional coordinates by applying furniture placement information included in structured design data to the generated internal space partition model. The internal modeling unit (142) identifies the type of furniture included in the furniture placement information and loads a 3D model of the furniture from a pre-stored furniture 3D model library (not shown).

[0201] The internal modeling unit (142) stores 3D models of various furniture, including tables, chairs, sofas, beds, cabinets, countertops, refrigerators, and air conditioners, in a furniture 3D model library. The furniture 3D model library stores actual size, material, and color information for each furniture 3D model so that they can be used immediately upon loading, and the furniture 3D models are stored in FBX, OBJ, and GLTF formats to ensure compatibility with 3D graphics software. The furniture 3D model library stores various design variations for each type of furniture so that the user can select a furniture design according to their preference, and each design variation is distinguished by a unique identification number.

[0202] The internal modeling unit (142) places the loaded furniture 3D model at the center coordinates included in the furniture placement information. The internal modeling unit (142) performs rotation and scaling for each placed furniture object. The internal modeling unit (142) rotates the furniture object around the z-axis by applying the rotation angle included in the furniture placement information, and transforms all vertex coordinates of the furniture object by applying a rotation transformation matrix. The internal modeling unit (142) adjusts the size of the furniture object by applying the width, length, and height included in the furniture placement information, and transforms the size of the furniture object by applying a scale transformation matrix.

[0203] The internal modeling unit (142) can maintain the width, length, and height ratios of the furniture 3D model when scaling, or ignore the ratios and transform it to a specified size. If the ratio maintenance flag is set in the furniture placement information, the internal modeling unit (142) maintains the ratio by applying the largest change rate among the width, length, and height equally to all axes, and if the ratio maintenance flag is not set, it transforms it to a specified size by applying an independent scale to each axis. The internal modeling unit (142) adjusts the z-coordinate of the lowest vertex of the furniture object to 0 so that the furniture object touches the floor exactly.

[0204] The internal modeling unit (142) creates a furniture placement model when the placement, rotation, and scaling of all furniture objects are completed. The internal modeling unit (142) transmits the furniture placement model to the rendering unit (144) and includes identification information, position coordinates, rotation angle, and size information for each furniture object in the furniture placement model. In the connection between the internal modeling unit (142) and the rendering unit (144), the internal modeling unit (142) notifies the rendering unit (144) as soon as it completes the furniture placement, and upon receiving the notification, the rendering unit (144) begins texture mapping and lighting processing.

[0205] The external modeling unit (143) receives a three-dimensional space basic model generated from the space modeling unit (141). The external modeling unit (143) identifies the exterior wall area of ​​the three-dimensional space basic model, and the exterior wall area is defined as a wall surface facing the outside of the building, and generally, when the rental space is located on the first floor or lower floor of the building, the wall surface facing the road corresponds to the exterior wall area.

[0206] The external modeling unit (143) extracts external finishing material information, signage information, and entrance information included in the structured design data transmitted from the data conversion unit (133). The external modeling unit (143) applies the finishing material type, color, and texture image included in the external finishing material information to the exterior wall area. The external modeling unit (143) expresses the material texture of the finishing material by performing texture mapping on the polygon mesh of the exterior wall, and maps the texture image to the exterior wall surface based on UV coordinates. The external modeling unit (143) expresses the three-dimensional effect of the surface by additionally applying a normal map and a displacement map depending on the type of finishing material.

[0207] The external modeling unit (143) creates a signboard model based on the location, size, shape, and color of the signboard included in the signboard information. The external modeling unit (143) identifies whether the shape of the signboard is a rectangle, a circle, an ellipse, or a polygon, and creates a polygon mesh of the corresponding shape. The external modeling unit (143) sets the width and height of the signboard model by applying the size included in the signboard information, and sets the thickness of the signboard to a standard value (e.g., 5cm to 10cm) or a value specified in the design data.

[0208] The external modeling unit (143) applies colors included in the sign information to the surface of the sign model, and if there is text information to be displayed on the sign, it creates the text as a 3D text object and places it on the surface of the sign. The external modeling unit (143) sets the font, size, and color of the text according to the sign information, and can express the text as protruding from the surface of the sign or engraved. If the sign includes lighting effects (LED sign, neon sign, etc.), the external modeling unit (143) applies an emissive material to the surface of the sign to express the effect of the sign emitting light.

[0209] The external modeling unit (143) places the generated signboard model at a designated location on the exterior wall. The external modeling unit (143) uses location coordinates included in the signboard information to attach the signboard model to the surface of the exterior wall or to place it at a certain distance from the exterior wall. If the signboard is of a protruding type, the external modeling unit (143) places the signboard model so that it protrudes forward from the exterior wall, and further creates a bracket model that supports the signboard and places it between the signboard and the exterior wall to enhance realism.

[0210] The external modeling unit (143) creates an entrance model using the location, size, and shape of the entrance included in the entrance information. The external modeling unit (143) processes the polygon mesh of the entrance area as an opening and places the entrance frame and door model. The external modeling unit (143) applies an appropriate shape among a glass door, a hinged door, and a sliding door to the door model, and the door shape uses information specified in the design data or industry-specific standard shapes. In the case of a glass door, the external modeling unit (143) applies a transparent material so that the interior is visible, and in the case of a hinged door or a sliding door, additionally models detailed elements such as door handles and hinges.

[0211] The external modeling unit (143) creates an external model to which the external finishing material, signboard, and entrance are all applied. The external modeling unit (143) transmits the created external model to the rendering unit (144) and includes the exterior wall area coordinates, finishing material information, signboard object information, and entrance object information in the external model. In the connection between the external modeling unit (143) and the rendering unit (144), the external modeling unit (143) notifies the rendering unit (144) as soon as it completes the creation of the external model, and the rendering unit (144) performs final rendering by integrating the internal model and the external model.

[0212] The rendering unit (144) receives a furniture placement model generated from the interior modeling unit (142) and an exterior model generated from the exterior modeling unit (143). The rendering unit (144) integrates the received models to form a complete three-dimensional interior model and generates a complete three-dimensional model that includes interior space partitioning, furniture placement, exterior finishing, and signage.

[0213] The rendering unit (144) performs texture mapping according to the finishing material information included in the structured design data transmitted from the data conversion unit (133). The rendering unit (144) maps the texture image specified in the finishing material information to each area of ​​the wall, floor, and ceiling. When mapping the texture, the rendering unit (144) calculates UV coordinates so that the texture image is applied to the surface without distortion, and the UV coordinates are expressed as 2D coordinates normalized to a range of 0 to 1 for each polygon vertex. The rendering unit (144) adjusts the tiling settings of the texture image so that the texture naturally fills the entire surface.

[0214] The rendering unit (144) sets the material properties of the finishing material as reflectivity, roughness, metallicity, and opacity. For example, the rendering unit (144) sets the marble finishing material to reflectivity 0.8, roughness 0.1, metallicity 0.0, and opacity 1.0, the wood finishing material to reflectivity 0.3, roughness 0.6, metallicity 0.0, and opacity 1.0, and the fabric finishing material to reflectivity 0.1, roughness 0.9, metallicity 0.0, and opacity 1.0. The rendering unit (144) inputs these material properties into a physically based rendering (PBR) model to accurately simulate the optical properties of the actual material.

[0215] The rendering unit (144) places a light source in a three-dimensional space based on the type of light, position of light, brightness of light, and color temperature information according to the lighting information included in the structured design data. The rendering unit (144) selects an appropriate light source type among a point light, a spotlight, a directional light, and an area light as the type of light. The rendering unit (144) uses a point light as a light source that emits light in all directions from a single point, such as a light bulb; uses a spotlight as a light source that emits light in a cone shape in a specific direction; uses a directional light as a light source that emits parallel light, such as sunlight; and uses an area light as a surface light source that provides a soft lighting effect.

[0216] The rendering unit (144) sets the three-dimensional coordinates of each light source according to the lighting position information. The rendering unit (144) receives lighting brightness information in units of lumens and sets the intensity of the light source, and converts the intensity of the light source into a unit used by the rendering engine by dividing the lumens value by a conversion factor according to the light source type. The rendering unit (144) receives color temperature information in units of Kelvin and sets the color of the light source, and converts the color temperature into RGB colors and applies them to the light source. For example, the rendering unit (144) converts a color temperature of 3000K into warm white light of RGB (255, 180, 107) and a color temperature of 6000K into cool white light of RGB (255, 244, 229).

[0217] The rendering unit (144) calculates the reflection, shadow, and brightness of light generated from the placed light source. The rendering unit (144) simulates the path of light using a ray tracing algorithm or a radiosity algorithm and calculates the reflection and refraction of light at each surface. When using the ray tracing algorithm, the rendering unit (144) backtracks the rays from the virtual camera toward each pixel to calculate the color and brightness of the surfaces where the rays meet, and recursively tracks the rays reflected or refracted from the surfaces to reflect the influence of indirect light. When using the radiosity algorithm, the rendering unit (144) calculates the transfer of light energy between each surface to implement a global illumination effect.

[0218] The rendering unit (144) uses a shadow mapping technique to generate shadows. The rendering unit (144) renders the scene from the viewpoint of each light source to generate a depth map, and during the final rendering, determines the shadow area by referencing the depth map and determines whether the corresponding pixel receives light directly from the light source or is shadowed by being obscured by another object. The rendering unit (144) applies a Percentage Closer Filtering (PCF) technique to express the boundaries of the shadows smoothly.

[0219] The rendering unit (144) generates a final 3D virtual interior model by performing photorealistic rendering. The rendering unit (144) accurately simulates the physical properties of the material using a physically based rendering (PBR) technique. The rendering unit (144) applies global illumination to reflect the influence of indirect light and calculates the effect of light reflecting from a surface and illuminating another surface to create a more realistic lighting environment. The rendering unit (144) applies ambient occlusion to emphasize the shading of corners and simulates the effect of ambient light being blocked in areas where two surfaces are close together.

[0220] By performing photorealistic rendering, the rendering unit (144) generates a 3D virtual interior model that is so realistic that it is difficult to distinguish it from a real photograph. Through this, building owners and customers can verify the interior results very accurately before actual construction, and can prevent dissatisfaction with results different from expectations after construction. Experimental results showed that more than 95% of building owners who received photorealistic rendering responded that they could accurately predict the final construction results by looking at the 3D virtual interior model, and the rate of requests for modification after construction decreased by about 70% compared to cases where only existing 2D blueprints were provided. This means that photorealistic rendering significantly improves the accuracy of decision-making and contributes to reducing modification costs after construction.

[0221] The rendering unit (144) generates image data of the final three-dimensional virtual interior model taken from multiple viewpoints. The rendering unit (144) places virtual cameras at various locations in a three-dimensional space and renders the three-dimensional virtual interior model at each location to save it as an image. The rendering unit (144) generates images from, for example, a front viewpoint (viewpoint looking inside from the entrance), a side viewpoint (viewpoint looking from the side of the space), a top viewpoint (viewpoint looking down from the ceiling), and a diagonal viewpoint (viewpoint looking diagonally from the corner of the space). The rendering unit (144) generates images from each viewpoint in high resolution (e.g., 1920×1080 pixels or higher) so that details are clearly visible.

[0222] The rendering unit (144) generates 360-degree panoramic image data. The rendering unit (144) places a virtual camera in the center of a three-dimensional space and generates a spherical panoramic image by rendering in all directions (up, down, left, right). To generate the spherical panoramic image, the rendering unit (144) renders images with a 90-degree field of view in each of the six directions (front, back, left, right, up, and down), combines them into a cubemap, and then converts them using an equiectangular projection method to save them as a two-dimensional image. The rendering unit (144) generates the 360-degree panoramic image in high resolution (e.g., 4096×2048 pixels or higher) so that the quality does not degrade when used in virtual reality mode.

[0223] The rendering unit (144) transmits the generated final 3D virtual interior model data, multiple viewpoint image data, and 360-degree panoramic image data to the model receiving unit (151) of the visualization unit (150). The rendering unit (144) saves and transmits the 3D virtual interior model data in one of the FBX, OBJ, or GLTF formats, and saves and transmits the image data in JPG or PNG formats. When saving in JPG format, the rendering unit (144) sets the compression quality to 90% or higher to minimize image quality degradation. In the connection between the rendering unit (144) and the visualization unit (150), a streaming method is used for large-capacity data transmission, and the rendering unit (144) divides the data into chunks and transmits them, and the visualization unit (150) receives and combines them sequentially.

[0224] When the rendering unit (144) receives a modification request from the modification processing unit (154) of the visualization unit (150), it updates the 3D virtual interior model by reflecting the modification content, re-renders the updated model, and transmits it to the visualization unit (150). If the modification request is a change in finishing material, lighting, or color, the rendering unit (144) re-renders only the relevant part to reduce processing time, and if the modification request is a change in furniture placement, it updates the furniture position in the internal modeling unit (142) and then re-renders the entire model. In the connection between the rendering unit (144) and the modification processing unit (154), a rendering priority queue is used to minimize the response time for the modification request, and the re-rendering work based on the modification request is processed before the new model creation work.

[0225] The AI ​​processing unit (145) includes a model storage unit (145a) and a model calculation unit (145b).

[0226] The model storage member (145a) stores a pre-trained second artificial intelligence model (AI2). The second artificial intelligence model (AI2) is a generative artificial intelligence model that has learned past interior design drawing data and three-dimensional model data corresponding to the design drawing.

[0227] The second artificial intelligence model (AI2) has a Generative Adversarial Network (GAN) or Variational Autoencoder (VAE) structure. In one embodiment of the present invention, the second artificial intelligence model (AI2) adopts a Conditional GAN ​​structure and is composed of a Generator network and a Discriminator network.

[0228] The generator network of the AI2 model has an encoder-decoder structure; the encoder receives a 2D blueprint image as input and compresses it into a latent vector, while the decoder generates 3D spatial coordinates, object position and size, and material and color information from the latent vector. The encoder consists of five convolution layers, each using a 3×3 kernel, stride of 2, and padding of 1 to progressively reduce the spatial resolution of the input image while extracting features. The decoder consists of five transposed convolution layers and reconstructs 3D information from the latent vector.

[0229] The discriminator network of the AI2 model receives 3D model data generated by the generator and actual 3D model data as input, and determines whether the data is real or generated. The discriminator network consists of 6 convolutional layers, and the final output is a probability value of 0 (fake) or 1 (real).

[0230] The training data for the second artificial intelligence model (AI2) consists of 25,000 interior projects collected over the past seven years, and for each project, a pair of 2D design images (DWG files converted into images or PDF designs) and 3D model data (FBX or OBJ format) created based on those designs are paired. The training data includes interiors for various uses, such as restaurants, cafes, offices, and residential spaces, enabling the model to learn various types of spaces.

[0231] The AI2 model is trained using supervised learning. During the training process, the generator creates a 3D model from a 2D blueprint image, and the discriminator learns to distinguish between the generated 3D model and the actual 3D model. The generator and the discriminator are trained adversarially, and the generator is improved to create a more sophisticated 3D model to deceive the discriminator, while the discriminator is improved to distinguish between the generated model and the actual model more accurately.

[0232] The loss function of the AI2 model consists of a generator loss function and a discriminator loss function. The generator loss function is defined as follows:

[0233] L_G = L_adv + λ_1 × L_rec + λ_2 × L_perc

[0234] Here, L_adv is the Adversarial Loss, which is a loss to deceive the discriminator; L_rec is the Reconstruction Loss, which is a loss to make the generated 3D model similar to the actual 3D model; and L_perc is the Perceptual Loss, which is a loss to make the high-level features of the generated 3D model similar to the actual model. λ_1 and λ_2 are the weights of each loss item, and in one embodiment of the present invention, λ_1 = 10 and λ_2 = 1.

[0235] The discriminator loss function is defined as follows:

[0236] L_D = -E[log(D(x_real))] - E[log(1 - D(G(z)))]

[0237] Here, D represents the discriminator network, G represents the generator network, x_real represents the actual 3D model data, z represents the 2D blueprint image, and E represents the expected value.

[0238] The Adam optimizer is used to train the AI2 model, and the learning rate is set to 0.0002 for both the generator and the discriminator. The AI2 model is trained for 200 epochs, and mini-batch gradient descent is applied with a batch size of 16 in each epoch. To prevent overfitting during the training process, the AI2 model applies dropout (dropout rate 0.5) and batch normalization.

[0239] After the training of the second AI model (AI2) is completed, the performance of the model is evaluated using a validation dataset (5,000 interior projects not used for training). The 3D coordinate prediction accuracy of the second AI model (AI2) was measured at 5.2 cm based on Mean Absolute Error (MAE), the object location and size prediction accuracy was measured at 8.7 cm based on MAE, and the material and color classification accuracy was measured at 92.3%. These prediction accuracy levels are sufficient for generating 3D models, and the second AI model (AI2) continuously improves its prediction accuracy by being retrained regularly (every 6 months) by adding new interior project data.

[0240] By utilizing the AI2 model, information not specified in the 2D blueprint can be inferred and automatically supplemented by the AI. For example, if height information for furniture is missing from the 2D floor plan, the AI2 model can estimate an appropriate height based on the type of furniture and its dimensions on the floor plan. Additionally, if the material of the wall is not specified, the AI2 model can recommend an appropriate material by considering the use of the space and surrounding elements. Experimental results showed that when using the AI2 model, a 3D model could be generated with over 90% accuracy even from incomplete blueprints, and work efficiency was significantly improved as the number of times the interior design company had to input additional information decreased by approximately 60%.

[0241] The model storage unit (145a) stores the weight parameters of the second artificial intelligence model (AI2) in a file format (e.g., HDF5, PyTorch .pth, TensorFlow .pb) and loads the second artificial intelligence model (AI2) into memory upon the request of the model computation unit (145b). The model storage unit (145a) performs version management of the second artificial intelligence model (AI2), retrains and updates the model whenever new training data is added, and records the version number, training date, and performance indicators for each version. The model storage unit (145a) stores multiple versions of the second artificial intelligence model (AI2) simultaneously so that it can be rolled back to a previous version as needed.

[0242] The model operation unit (145b) inputs the structured design data transmitted from the data conversion unit (133) into the second artificial intelligence model (AI2). The model operation unit (145b) preprocesses the wall information, furniture arrangement information, finishing material information, and lighting information included in the structured design data to match the input format of the second artificial intelligence model (AI2), converts it into a 2D design image, or converts the structured data in JSON format into a tensor format.

[0243] The model computation member (145b) executes the second artificial intelligence model (AI2) to generate three-dimensional coordinate information, object placement information, and texture information. The model computation member (145b) derives the exact location of each wall and furniture from the three-dimensional coordinate information output by the second artificial intelligence model (AI2), derives the rotation angle and size of each object from the object placement information, and derives the texture image and material properties to be applied to each surface from the texture information. The model computation member (145b) checks the reliability score included in the output of the second artificial intelligence model (AI2) and selectively uses only the information with high reliability.

[0244] The model computation unit (145b) transmits the generated 3D coordinate information, object placement information, and texture information to the spatial modeling unit (141), internal modeling unit (142), external modeling unit (143), and rendering unit (144), respectively. The model computation unit (145b) selectively transmits only the information required by each unit, transmitting the 3D coordinate information to the spatial modeling unit (141), the object placement information to the internal modeling unit (142) and external modeling unit (143), and the texture information to the rendering unit (144). In the connection between the model computation unit (145b) and each modeling unit, the model computation unit (145b) transmits the information to each modeling unit asynchronously as soon as it is generated, so that each modeling unit can perform processing in parallel.

[0245] The model operation member (145b) controls the structured design data to prioritize the application of information if there is information explicitly specified in the structured design data, and to utilize information generated by the second artificial intelligence model (AI2) as auxiliary information if there is no explicit information. For example, if the location coordinates of a specific piece of furniture are specified in the structured design data, the model operation member (145b) uses those coordinates as they are, and if the location coordinates are not specified, it uses the location coordinates estimated by the second artificial intelligence model (AI2). When merging explicit information and artificial intelligence-generated information, the model operation member (145b) sets a "Confirmed" flag on the explicit information and sets an "Estimated" flag on the artificial intelligence-generated information to clearly distinguish the source of the information.

[0246] The model computation unit (145b) evaluates the reliability of information generated by the second artificial intelligence model (AI2). The model computation unit (145b) checks the reliability score (value between 0 and 1) included in the output of the second artificial intelligence model (AI2) and uses the information only if the reliability score is greater than or equal to a preset threshold (e.g., 0.7). If the reliability score is less than the threshold, the model computation unit (145b) does not use the information and applies a default value, or sends a request for additional information input to the representative terminal (P3) of the interior company. The model computation unit (145b) records the reliability evaluation results in a log to monitor the performance of the second artificial intelligence model (AI2) and utilizes it for model improvement.

[0247] The visualization unit (150) provides the 3D virtual interior model generated from the model generation unit (140) to the building owner's terminal (P1) or the customer's terminal (P2), receives a modification request, and transmits it to the model generation unit (140). The visualization unit (150) visualizes the 3D virtual interior model so that the user can intuitively check it, and provides an interactive modification function so that the user can adjust the interior in real time and check the final result. The visualization unit (150) acts as an interface between the model generation unit (140) and the user terminal, and converts complex 3D model data into a user-friendly form so that even non-expert building owners and customers can easily understand the interior results and make decisions.

[0248] The visualization unit (150) includes a model receiving unit (151), a model providing unit (152), a modification request input unit (153), and a modification processing unit (154), and these components sequentially process data and are responsible for visualizing the 3D model, user interaction, and processing modification requests.

[0249] The model receiving unit (151) receives the final 3D virtual interior model data, multiple viewpoint image data, and 360-degree panoramic image data transmitted from the rendering unit (144). The model receiving unit (151) checks the file format of the received data, verifies file integrity, and analyzes the file header to check whether the data is not damaged. In the connection between the model receiving unit (151) and the rendering unit (144), a chunk-based streaming protocol is used for large-capacity data transmission, and the rendering unit (144) divides the data into chunks of a certain size and transmits them, and the model receiving unit (151) receives the chunks sequentially and combines them to distribute the network load.

[0250] The model receiving unit (151) converts the received data into a format that can be transmitted to the building owner's terminal (P1) or the customer's terminal (P2). The model receiving unit (151) converts the 3D virtual interior model data into a WebGL format that can be rendered in a web browser or a lightweight 3D format optimized for mobile environments (e.g., glTF 2.0). When converting to the WebGL format, the model receiving unit (151) generates model data in JSON format compatible with the Three.js library and constructs complete 3D model data including vertex coordinates, normal vectors, UV coordinates, and material properties.

[0251] The model receiving unit (151) converts image data into a compression format suitable for network transmission and adjusts the resolution to match the screen size of the user's terminal. The model receiving unit (151) receives screen resolution information from the user's terminal and, if the image resolution exceeds 1.5 times the screen resolution, downsamples the image to reduce the file size. When converting to JPG format, the model receiving unit (151) sets the compression quality to a range of 85-95% to balance visual quality and file size. The model receiving unit (151) improves the user experience by using the Progressive JPEG format during the conversion process so that the image is loaded progressively.

[0252] The model receiving unit (151) transmits the converted data to the model providing unit (152). The model receiving unit (151) sets a compression algorithm to minimize the loss of quality of the original data during the data conversion process and uses a combination of lossy compression and lossless compression. The model receiving unit (151) maintains accuracy by applying lossless compression to the geometric structure (vertices, faces) of the 3D model, and reduces the file size by applying lossy compression to texture images. In the connection between the model receiving unit (151) and the model providing unit (152), the converted data is stored in shared memory, and the model providing unit (152) references the corresponding memory address to eliminate data copy overhead.

[0253] The model providing unit (152) transmits the converted 3D virtual interior model data from the model receiving unit (151) to the building owner's terminal (P1) or the customer's terminal (P2). The model providing unit (152) transmits the data using the HTTP / 2 protocol and reduces the transmission time by utilizing the multiplexing function to transmit the 3D model data and image data simultaneously. The model providing unit (152) monitors the network status during transmission to optimize the transmission speed and automatically adjusts the data quality when the network bandwidth is narrow to reduce the loading time.

[0254] The model providing unit (152) enables a 3D virtual interior model to be displayed on the display unit of the building owner's terminal (P1) or the customer's terminal (P2). The model providing unit (152) provides a web-based 3D viewer or a 3D viewer in the form of a native application, and enables the user to view the 3D virtual interior model through a browser or app. When the model providing unit (152) provides a web-based 3D viewer, it renders the 3D model using HTML5 Canvas elements and WebGL API, and optimizes rendering performance by utilizing a 3D graphics library such as Three.js or Babylon.js.

[0255] The model providing unit (152) provides an interactive viewer function that enables rotation, zooming in and out, and viewpoint changes for the displayed 3D virtual interior model. The model providing unit (152) rotates the viewpoint of the 3D virtual interior model up, down, left, and right in response to a touch from the building owner's terminal (P1) or the customer's terminal (P2). The model providing unit (152) detects a touch drag gesture, calculates the difference in coordinates between the drag start point and the end point, converts this into a camera rotation angle, maps horizontal drags to y-axis rotation (Yaw), and vertical drags to x-axis rotation (Pitch). The model providing unit (152) applies a sensitivity coefficient when converting the rotation angle so that sufficient rotation is achieved even with small drag movements, and the sensitivity coefficient can be adjusted according to user settings.

[0256] The model providing unit (152) performs zooming in or out in response to pinch gestures or scroll inputs. When the model providing unit (152) detects a pinch-out gesture of spreading two fingers, it calculates the change in distance between the two fingers and zooms in by moving the camera closer to the 3D model according to the rate at which the distance increases. When the model providing unit (152) detects a pinch-in gesture of bringing two fingers together, it zooms out by moving the camera further away from the 3D model. The model providing unit (152) detects mouse scroll inputs to provide the same zoom in and zoom out functions, mapping scroll up to zoom in and scroll down to zoom out. The model providing unit (152) adjusts the distance between the camera and the center of the model during zooming in and out, and sets a minimum distance and a maximum distance to prevent excessive zooming in or out.

[0257] The model providing unit (152) moves the camera to the corresponding viewpoint when the user clicks a predefined viewpoint button. The model providing unit (152) provides front viewpoint, side viewpoint, top viewpoint, and diagonal viewpoint buttons, and when each button is clicked, the camera automatically moves to the corresponding position and angle. The model providing unit (152) applies tween animation when moving the camera to provide a smooth transition, and by moving gradually from the current camera position to the target position over 0.5 to 1 second, it prevents user confusion caused by abrupt changes in viewpoint.

[0258] Through the interactive viewer function provided by the model provider (152), users can freely explore and view the 3D virtual interior model from various angles. Through this, users can grasp the overall structure and detailed features of the space, and intuitively understand spatial awareness, movement patterns, and views that are difficult to verify in 2D blueprints. Experimental results showed that 98% of users who used the interactive viewer function responded that they could fully understand the space through the 3D model, and the level of understanding improved by about 50% compared to when only 2D blueprints were provided (65%). In addition, when the interactive viewer function was used, the rate at which users gave final approval with confidence in the interior results increased by about 40%, which improved the speed of project progress and the quality of decision-making.

[0259] The model providing unit (152) provides a virtual reality mode using 360-degree panoramic image data transmitted from the model receiving unit (151). The model providing unit (152) allows the user to freely look around the interior of a three-dimensional virtual interior model through a display in the virtual reality mode. The model providing unit (152) maps the 360-degree panoramic image onto a sphere mesh, renders the corresponding area according to the user's gaze direction, and when the user moves the terminal or drags the screen, the gaze direction changes and a different part of the panoramic image is displayed.

[0260] The model providing unit (152) can provide an immersive virtual reality experience by linking with a virtual reality headset (e.g., Oculus Quest, HTC Vive). The model providing unit (152) receives gyroscope and accelerometer sensor data from the headset and changes the viewpoint in real time according to the user's head movement, and applies head tracking technology so that the scene being viewed changes as the user turns their head. The model providing unit (152) provides a sense of depth by implementing binocular disparity in virtual reality mode, and enhances the sense of three-dimensional space by rendering an image for the left eye and an image for the right eye separately.

[0261] The modification request input unit (153) receives modification request information including at least one of a furniture arrangement change request, a finishing material change request, a lighting change request, and a color change request from the building owner's terminal (P1) or the customer's terminal (P2). The modification request input unit (153) receives a user selecting a specific object on a 3D virtual interior model displayed on the building owner's terminal (P1) or the customer's terminal (P2), and receives at least one modification command for the selected object, such as moving, rotating, replacing, or deleting. A real-time event communication channel is maintained between the modification request input unit (153) and the model providing unit (152) so that all user inputs are immediately transmitted to the modification request input unit (153).

[0262] When a user clicks or touches a specific piece of furniture on a 3D virtual interior model, the modification request input unit (153) changes the furniture object to a selected state and highlights it. The modification request input unit (153) uses a ray casting technique for object selection, emits a ray from the screen coordinates clicked by the user into 3D space, and determines the first object that intersects the ray as the selected object. The modification request input unit (153) applies an outline effect to the selected object to highlight it, and uses a bright color (e.g., yellow, cyan) that contrasts with the surrounding colors for the outline color to ensure visibility.

[0263] The modification request input section (153) displays move, rotate, replace, and delete buttons for the selected furniture object, allowing the user to click the corresponding buttons to input modification commands. When the move button is clicked, the modification request input section (153) changes the furniture object to a draggable state, allowing the user to drag the furniture to move it to a new location, and the furniture object moves along with the mouse cursor or finger during movement. When the rotate button is clicked, the modification request input section (153) displays a rotation handle around the furniture object, allowing the user to drag the rotation handle to rotate the furniture.

[0264] When the modification request input unit (153) receives a request to change the furniture arrangement, it detects a gesture in which the user drags the selected furniture to move it to a new location. The modification request input unit (153) sets the coordinates at the time the drag ends as the new location coordinates and performs an inverse projection calculation to convert the screen coordinates into three-dimensional space coordinates. The modification request input unit (153) automatically adjusts the z-coordinate so that the furniture is placed on the floor plane (z=0), performs collision detection to check if the furniture overlaps with a wall or other furniture, and if a collision is detected, returns the furniture to its previous position or displays a warning message.

[0265] When the modification request input unit (153) receives a request to change the finishing material, it allows the user to select a wall or floor area to be changed and displays a list of finishing materials for the selected area. The modification request input unit (153) provides various finishing material options such as marble, wood, tile, wallpaper, and paint in the list of finishing materials, and displays thumbnail images for each finishing material so that the user can visually select the finishing material. When the user selects the desired finishing material from the list of finishing materials, the modification request input unit (153) includes the texture image path and material attributes of the corresponding finishing material in the modification request information.

[0266] When the modification request input unit (153) receives a request to change lighting, it allows the user to select a lighting object to be changed and displays a lighting brightness slider and a color temperature slider for the selected lighting object. When the user adjusts the brightness slider, the modification request input unit (153) converts the slider value into a brightness value in lumens, and when the user adjusts the color temperature slider, it converts the slider value into a color temperature value in Kelvin and includes it in the modification request information. The modification request input unit (153) displays a real-time preview of the lighting effect while the user is adjusting the sliders, so that the user can immediately check the adjustment result.

[0267] When the modification request input unit (153) receives a color change request, it allows the user to select an object to be changed and displays a color palette for the selected object. The modification request input unit (153) provides basic colors (red, orange, yellow, green, blue, purple, white, gray, black) and user-defined colors in the color palette, and when the user selects a desired color from the color palette, it includes the RGB or HEX code of the corresponding color in the modification request information. The modification request input unit (153) provides a real-time preview when a color is selected so that the color of the selected object is changed and displayed immediately.

[0268] The modification request input unit (153) analyzes the input modification request information to extract information on the object to be modified and information on the modification content. The modification request input unit (153) extracts the object's unique identification number, object type (furniture, wall, lighting, etc.), current location coordinates, current rotation angle, current size, current color, and current material as information on the object to be modified, and extracts the relevant information among the location coordinates, rotation angle, size, texture, color, brightness, and color temperature to be changed as information on the modification content. The modification request input unit (153) clearly distinguishes and records the state before modification and the state after modification, thereby enabling the modification processing unit (154) to accurately track the modification history.

[0269] The modification request input unit (153) transmits the extracted modification target object information and modification content information to the modification processing unit (154). When modification requests occur simultaneously for multiple objects, the modification request input unit (153) organizes the modification information for each object into an array and transmits it, and sets the priority of the modification requests so that important modification requests are processed first. A modification request queue is used in the connection between the modification request input unit (153) and the modification processing unit (154), and the modification request input unit (153) adds modification requests to the queue, and the modification processing unit (154) retrieves modification requests from the queue and processes them sequentially, thereby stably processing simultaneous modification requests.

[0270] The modification processing unit (154) transmits the modification target object information and modification content information extracted from the modification request input unit (153) to the model generation unit (140). If the modification request information is a request to change the furniture placement, the modification processing unit (154) transmits the identification information of the furniture object to be modified and the location coordinate information to be changed to the internal modeling unit (142), and the internal modeling unit (142) updates the location of the furniture object and performs re-rendering. In the connection between the modification processing unit (154) and the internal modeling unit (142), a Remote Procedure Call (RPC) method is used so that the modification processing unit (154) remotely calls the furniture location update function of the internal modeling unit (142) and receives the result.

[0271] When the modification request information is a request to change the exterior finishing material or a request to change the signboard, the modification processing unit (154) transmits the information on the exterior wall area to be modified and the information on the finishing material or signboard to be changed to the exterior modeling unit (143). When the modification request information is a request to change the finishing material, a request to change the lighting, or a request to change the color, the modification processing unit (154) transmits the information on the area of ​​the wall or floor to be modified and the texture information of the finishing material to be changed to the rendering unit (144), and the rendering unit (144) updates the texture or material properties of the area and performs re-rendering. In the connection between the modification processing unit (154) and the rendering unit (144), a partial rendering technique is applied to selectively re-render only the modified area, thereby reducing the processing time by about 70% compared to re-rendering the entire model.

[0272] The modification processing unit (154) receives the modified final 3D virtual interior model data from the rendering unit (144). The modification processing unit (154) provides the received modified final 3D virtual interior model data to the building owner's terminal (P1) or the customer's terminal (P2) via the model receiving unit (151) and the model providing unit (152). When the modified model is displayed to the user, the modification processing unit (154) allows the user to check the modification results and select additional modification or approval, and provides a modification confirmation button and an additional modification button. When the user clicks the modification confirmation button, the modification processing unit (154) finally confirms the modification, and when the user clicks the additional modification button, it returns control to the modification request input unit (153) to receive an additional modification request.

[0273] The modification processing unit (154) generates modification history data including modification request time information, modification content information, and modification count information. The modification processing unit (154) records the date and time when the modification request occurred in UTC (Coordinated Universal Time) format as modification request time information, and records the modification target object information, pre-modification state information, and post-modification state information as modification content information. The modification processing unit (154) records the cumulative number of modifications that occurred for the same 3D virtual interior model as modification count information, and increases the modification count by 1 each time a modification request occurs.

[0274] The modification processing unit (154) includes modification request time information, modification target object information, pre-modification state information, post-modification state information, and modification requester information in the modification history data. The modification processing unit (154) records identification information (user ID, name, email) of the building owner or customer as modification requester information, thereby enabling tracking of who requested which modification. The modification processing unit (154) provides an option to input the reason for modification into the modification history data, and if the user inputs the reason for modification, it is included in the modification history data for future analysis and learning.

[0275] The modification processing unit (154) transmits the generated modification history data to the modification history storage unit (163) of the data management unit (160). When transmitting the modification history data, the modification processing unit (154) transmits the rental space identification information and the 3D virtual interior model identification information together so that the data management unit (160) can properly classify and store the modification history. In the connection between the modification processing unit (154) and the modification history storage unit (163), a transaction processing mechanism is applied so that the modification completion process is not processed until the modification history data is completely stored in the database, thereby ensuring data consistency.

[0276] The modification processing unit (154) analyzes the pattern of modification requests when modification requests are repeated three or more times for the same three-dimensional virtual interior model. To analyze the modification request pattern, the modification processing unit (154) statistically processes the type of modified object, the type of modification content, and the modification frequency, and automatically identifies the modification pattern using a machine learning algorithm (e.g., K-means clustering, decision tree). For example, if a specific piece of furniture has been moved three or more times or the finishing material of a specific wall has been changed three or more times, the modification processing unit (154) identifies the object as a frequently modified object.

[0277] The modification processing unit (154) extracts information on frequently modified objects and modification direction information for frequently modified objects. The modification processing unit (154) analyzes the trends of the user's preferred location, color, material, and lighting brightness as modification direction information. For example, if the user shows a pattern of continuously moving the table toward the window, it extracts a trend of "preferring the table → window direction." The modification processing unit (154) analyzes color change patterns to identify the user's tendency to change from bright colors to dark colors or from warm colors to cool colors.

[0278] The modification processing unit (154) transmits the extracted information to the model operation unit (145b) of the AI ​​processing unit (145) to be used as training data for the second artificial intelligence model (AI2). When converting the extracted information into a training data format, the modification processing unit (154) sets the design information before modification as input data and the design information after modification as output data, and generates an input-output pair so that the second artificial intelligence model (AI2) can learn the user's preferences. The modification processing unit (154) assigns weights to the training data, giving high weights to the training data of objects with a high number of modifications, thereby allowing the second artificial intelligence model (AI2) to learn important patterns first.

[0279] The modification processing unit (154) enables the second artificial intelligence model (AI2) to learn the user's preferences and generate a 3D model that matches the user's preferences in future similar projects. This allows the user's preferences to be reflected from the initial model generation stage, thereby reducing the number of modification requests and improving the speed of project progress. An asynchronous data transmission method is used in the connection between the modification processing unit (154) and the AI ​​processing unit (145), so that the modification processing unit (154) can proceed with the next task immediately after transmitting the training data, and the AI ​​processing unit (145) independently performs model retraining.

[0280] By having the modification processing unit (154) process modification requests in real time and immediately provide the modified results to the user, the user can try out various options before construction and select the optimal interior. This prevents additional costs and wasted time due to modifications after construction, and improves user satisfaction. As a result of the experiment, the average number of modification requests made by users who received the modification request function was 4.2, and it took an average of 2.3 days to final approval. On the other hand, when only a 2D design was provided without the modification request function, an average of 1.8 on-site modifications occurred after construction, and the modification costs accounted for about 15% of the total project cost. This means that the modification request function has the effect of reducing construction costs by about 15% and shortening the project completion time by about one week.

[0281] The data management unit (160) stores matching result data transmitted from the vendor selection unit (120), 3D virtual interior model data generated from the model generation unit (140), and modification history data transmitted from the visualization unit (150), and manages the entry history and interior progress status for each rental space. The data management unit (160) systematically manages the data of the entire project, tracks the progress status in real time, and supports continuous performance improvement of the system by analyzing the accumulated data and utilizing it as training data for the artificial intelligence model. The data management unit (160) acts as the data hub of the entire system and realizes the intelligence of the system by collecting, storing, and analyzing data generated from each unit.

[0282] The data management unit (160) includes a matching data storage unit (161), a model data storage unit (162), a modification history storage unit (163), a progress status management unit (164), and a training data generation unit (165), and these components each handle a specific type of data while interlocking to perform integrated data management.

[0283] The matching data storage unit (161) receives matching result data transmitted from the matching execution unit (124). The matching data storage unit (161) extracts rental space identification information, matched business identification information, and matching completion time information included in the received matching result data, and a message queue method is used in the connection between the matching data storage unit (161) and the matching execution unit (124) so ​​that the matching data storage unit (161) can process them sequentially and stably even if the matching execution unit (124) transmits multiple matching results simultaneously.

[0284] The matching data storage unit (161) stores the extracted information in the leased space-specific entry history database (DB4). The leased space-specific entry history database (DB4) is implemented as a relational database (e.g., MySQL, PostgreSQL) or a NoSQL database (e.g., MongoDB) and stores past and present entry history for each leased space in a time series. The leased space-specific entry history database (DB4) uses leased space identification information as a primary key and stores matched business identification information, matching completion time information, contract start time information, contract end time information, and whether entry was successful. The leased space-specific entry history database (DB4) applies an index to support fast querying based on leased space identification information and matching completion time information.

[0285] The matching data storage unit (161) adds a new record to the entry history database (DB4) for each rental space whenever new matching result data is received. If there is a previous matching history for the same rental space, the matching data storage unit (161) determines whether to re-match by comparing it with the previous history, and if it is a re-match, records the end time of the previous matching as the current time and creates a new matching record. To track whether entry is successful, the matching data storage unit (161) periodically checks whether business continues after entry by linking with an external system (e.g., the company's headquarters system, POS system), and records it as successful entry if business continues for more than 6 months.

[0286] The matching data storage unit (161) transmits the stored matching result data to the progress status management unit (164). The matching data storage unit (161) transmits building owner identification information and building owner contact information along with the matching result data, so that the progress status management unit (164) can notify the building owner of the progress status. In the connection between the matching data storage unit (161) and the progress status management unit (164), a Publish-Subscribe pattern is used so that when the matching data storage unit (161) publishes a new matching result, subscribers including the progress status management unit (164) are automatically notified.

[0287] The model data storage unit (162) receives three-dimensional virtual interior model data transmitted from the rendering unit (144). The model data storage unit (162) tags the received three-dimensional virtual interior model data with rental space identification information, interior company identification information, and model creation time information, and the tag information is stored in the header of the model data file or a separate metadata file in the format of metadata.

[0288] The model data storage unit (162) stores tagged 3D virtual interior model data in the interior model database (DB5). The interior model database (DB5) is implemented as a hybrid structure of a file-based storage (e.g., file system, object storage) and a metadata database (e.g., MySQL, PostgreSQL), and the 3D model file is stored in the file storage, while the model file path, model creation time information, and model version information are stored in the metadata database. The interior model database (DB5) stores the 3D virtual interior model created for each rental space, and stores the model file path, model creation time information, and model version information together.

[0289] The model data storage unit (162) stores 3D virtual interior model data in a file format (FBX, OBJ, GLTF) and records the storage path in the interior model database (DB5). The model data storage unit (162) organizes the file storage path into a hierarchical directory structure and classifies and stores files in the format of year / month / day / rental space ID / , thereby improving the efficiency of file management and searching. When multiple versions of a 3D virtual interior model are created for the same rental space, the model data storage unit (162) stores each version as a separate file and assigns a version number, with the version number increasing in the format v1, v2, v3, etc. The model data storage unit (162) supports the user in reviewing previous versions of the model or reverting to previous versions through version management.

[0290] The model data storage unit (162) transmits the stored 3D virtual interior model data to the progress status management unit (164). The model data storage unit (162) transmits rental space identification information, model creation time information, and model version information as 3D virtual interior model creation history data, and in the connection between the model data storage unit (162) and the progress status management unit (164), the model data storage unit (162) notifies the progress status management unit (164) as soon as it saves a new model, so that the progress status management unit (164) can update the interior progress stage to the "model creation stage."

[0291] The modification history storage unit (163) receives modification history data transmitted from the modification processing unit (154). The modification history storage unit (163) analyzes the modification request time information, modification target object information, and modification content information included in the received modification history data. In the connection between the modification history storage unit (163) and the modification processing unit (154), an acknowledgment mechanism is applied to ensure reliable message transmission, and when the modification history storage unit (163) successfully receives the modification history data, it transmits an acknowledgment to the modification processing unit (154).

[0292] The modification history storage unit (163) stores the analyzed information in the modification history database (DB6) for each rental space. The modification history database (DB6) is implemented as a time-series database (e.g., InfluxDB, TimescaleDB) or a relational database and stores all modification history for the 3D virtual interior model of each rental space in a time series. The modification history database (DB6) uses rental space identification information and 3D virtual interior model identification information as composite primary keys and stores modification request time information, modification target object information, state information before modification, state information after modification, modification requester information, and modification count information. The modification history database (DB6) creates an index based on the modification request time information to enable fast time-range-based querying.

[0293] The modification history storage unit (163) adds a new record to the modification history database (DB6) whenever new modification history data is received. When multiple modifications occur for the same object, the modification history storage unit (163) stores each modification as a separate record and records the number of modifications accumulated. The modification history storage unit (163) serializes and stores the state before modification and the state after modification in JSON format, thereby allowing for flexible storage and retrieval of complex object states.

[0294] The modification history storage unit (163) transmits the stored modification history data to the progress status management unit (164) and the training data generation unit (165). The modification history storage unit (163) statistically processes and transmits modification frequency information and major modification item information along with the modification history data. The modification frequency is calculated as the number of modification requests per unit of time (e.g., 1 day, 1 week), and the major modification items are identified as the most frequently modified object type and modification content type. In the connection between the modification history storage unit (163) and the training data generation unit (165), when the modification history storage unit (163) accumulates a certain number (e.g., 100) or more of modification history, it transmits it in batches to the training data generation unit (165) to trigger a training data generation operation.

[0295] The progress status management unit (164) calculates the number of matching attempts, the number of matching successes, and the average matching time for each rental space based on the store entry history data stored in the matching data storage unit (161). The progress status management unit (164) calculates the number of matching attempts by aggregating the number of records stored in the store entry history database (DB4) for each rental space, and calculates the number of matching successes by aggregating the number of records where the store entry success status is True. The progress status management unit (164) calculates the average matching time by averaging the elapsed time from the time of the matching request to the time of the matching completion for all matching history, and the elapsed time is expressed in units of hours or days.

[0296] The progress status management unit (164) determines the interior progress stage for each rental space based on the 3D virtual interior model creation history data stored in the model data storage unit (162). The progress status management unit (164) divides the interior progress stage into five stages: matching completion stage, design progress stage, model creation stage, modification progress stage, and final approval stage. Each stage proceeds sequentially, and when the completion condition of each stage is met, it transitions to the next stage.

[0297] When the progress status management unit (164) determines that the interior progress stage of the corresponding rental space is the matching completion stage when the matching result data is stored in the matching data storage unit (161). When the progress status management unit (164) receives a signal that the interior design drawing data reception is complete from the design data reception unit (132), it determines that the interior progress stage of the corresponding rental space is the design progress stage. An event communication channel is established between the design data reception unit (132) and the progress status management unit (164), so that when the design data reception unit (132) completes the reception of the design drawing data, it immediately notifies the progress status management unit (164).

[0298] When 3D virtual interior model data is stored in the model data storage unit (162), the progress status management unit (164) determines the interior progress stage of the corresponding rental space as the model creation stage. When modification history data is stored in the modification history storage unit (163), the progress status management unit (164) determines the interior progress stage of the corresponding rental space as the modification progress stage, and in the modification progress stage, the number of modification history data is displayed together so that the building owner can check how many modifications have been made so far.

[0299] When the progress status management unit (164) receives a final approval signal from the building owner's terminal (P1), it determines the interior progress stage of the leased space as the final approval stage. When the progress status management unit (164) receives the final approval signal, it records that the interior project of the leased space is completed, stores completion time information, and sends a final approval completion notification to the interior company's representative terminal (P3) to notify that construction work can begin.

[0300] The progress status management unit (164) provides the determined interior progress stage information to the building owner's terminal (P1). The progress status management unit (164) visually displays the current progress stage on the display unit of the building owner's terminal (P1) and indicates whether each stage is completed using a progress bar or a stage icon. The progress status management unit (164) divides the progress bar into five sections so that each section represents one stage, fills in the color to indicate completed stages, applies a blinking effect to stages in progress, and displays incomplete stages in gray.

[0301] The progress status management unit (164) sends a notification message to the building owner's terminal (P1) whenever the interior progress stage changes. The progress status management unit (164) includes information on the changed progress stage, the time of change, and the estimated completion time in the notification message. The progress status management unit (164) calculates the estimated completion time based on the average time required for similar past projects. For example, if it took an average of 5 days from the model creation stage to the final approval stage in past projects, the estimated completion time is 5 days after the current project enters the model creation stage. The progress status management unit (164) sends the notification message via a method selected by the building owner, such as a push notification, email, or SMS.

[0302] By having the progress status management unit (164) track the interior progress stages in real time and provide them to the building owner, the building owner can clearly understand the current status of the project and predict the completion time of each stage. Through this, the building owner can plan the move-in schedule in advance and can identify and respond to any delays early on. As a result of the experiment, 92% of building owners who received the progress status management function responded that they could clearly understand the progress of the project, and were able to recognize and take countermeasures within an average of 1.5 days when a project delay occurred. This means that the progress status management function has the effect of improving the transparency of the project, reducing uncertainty for building owners, and improving project management efficiency.

[0303] The training data generation unit (165) analyzes the modification history data stored in the modification history storage unit (163). The training data generation unit (165) queries all modification history stored in the modification history database (DB6) and classifies the modification history by rental space type.

[0304] The learning data generation unit (165) classifies rental space types into restaurants, cafes, unmanned stores, offices, and storage. The learning data generation unit (165) automatically classifies rental space types based on usage information of each rental space or business type information of tenants, and if usage information is unclear, it infers the rental space type by referring to the business type code of the tenant. The learning data generation unit (165) groups modification history data by classified type and performs analysis independently for each group.

[0305] The learning data generation unit (165) classifies objects that have been modified two or more times among the modification history data as frequently modified objects. The learning data generation unit (165) extracts state information after the final modification for the frequently modified objects and generates preferred design data. The learning data generation unit (165) includes object type, final location coordinates, final color, final material, and final size information in the preferred design data, and determines that this best reflects the user's preference as it is the design finally selected by the user after multiple modifications.

[0306] The learning data generation unit (165) statistically processes items that are frequently modified by type of rental space. The learning data generation unit (165) analyzes patterns such as, for example, that table arrangement is most frequently modified in restaurant type (35% of total modifications), lighting brightness is most frequently modified in cafe type (28% of total modifications), and partition location is most frequently modified in office type (40% of total modifications). The learning data generation unit (165) visualizes the results of the statistical processing in the form of a bar graph or a pie chart and provides them to the system administrator so that the modification patterns by type can be intuitively understood.

[0307] The learning data generation unit (165) inputs statistically processed data into a pre-trained third artificial intelligence model (AI3) to derive optimal interior design patterns for each rental space type. The third artificial intelligence model (AI3) is an artificial intelligence model that has learned past modification history data, final approved 3D virtual interior model data, and rental space type information.

[0308] The third artificial intelligence model (AI3) uses clustering algorithms (e.g., K-means, DBSCAN) and pattern recognition algorithms (e.g., Apriori, FP-Growth) to group similar modification patterns and derives the design elements that appear most frequently in each group as the optimal design pattern. The third artificial intelligence model (AI3) outputs recommended location coordinates and placement intervals for each piece of furniture as the optimal furniture placement pattern, outputs recommended color combinations to be applied to walls, floors, ceilings, and furniture as the optimal color combination, and outputs recommended location, brightness, and color temperature of lighting as the optimal lighting placement.

[0309] The third artificial intelligence model (AI3) is trained using an unsupervised learning method, and the training data consists of 8,000 modification history data collected over the past three years and the final approved 3D virtual interior model data for the project. The third artificial intelligence model (AI3) extracts features such as object type, location before modification, location after modification, color before modification, color after modification, and number of modifications from each modification history data, and performs clustering based on these features to group projects with similar modification patterns.

[0310] The third artificial intelligence model (AI3) extracts the most frequently occurring final state (location, color, size, etc.) in each cluster as the optimal design pattern for that cluster. For example, if the pattern in the cluster of restaurant types where tables are placed within 2 meters of a window, the wall color is beige, and the lighting color temperature is 3000K appears most frequently, the third artificial intelligence model (AI3) derives this as the optimal design pattern for the restaurant type.

[0311] The third artificial intelligence model (AI3) uses a pattern recognition algorithm to derive association rules between design elements. For example, the third artificial intelligence model (AI3) derives an association rule such as "if the wall color is white, the furniture color is dark at least 75% of the time," and includes this association rule in optimal design patterns to recommend harmonious color combinations.

[0312] The performance of the third AI model (AI3) is evaluated using a validation dataset (2,000 modification history data not used for training). When the design patterns recommended by the third AI model (AI3) were applied to the initial model, the number of user modification requests was measured to be an average of 1.8, which is a decrease of approximately 57% compared to when the third AI model (AI3) was not used (average 4.2). This means that the third AI model (AI3) recommends design patterns that accurately reflect user preferences.

[0313] The learning data generation unit (165) inputs the preferred design data collected for each type into the third artificial intelligence model (AI3) to derive the optimal furniture arrangement pattern, optimal color combination, and optimal lighting arrangement for each type. The learning data generation unit (165) transmits the derived optimal interior design pattern to the model operation unit (145b) of the AI ​​processing unit (145) to be used as training data for the second artificial intelligence model (AI2).

[0314] When the training data generation unit (165) generates training data for the second artificial intelligence model (AI2), it sets rental space type information and basic design information as input data and sets optimal interior design pattern information as output data. The training data generation unit (165) transmits the generated training data to the AI ​​processing unit (145), and the AI ​​processing unit (145) uses the received training data to retrain the second artificial intelligence model (AI2). A batch transmission method is used in the connection between the training data generation unit (165) and the AI ​​processing unit (145), so that when the training data generation unit (165) accumulates a certain number (e.g., 500) or more of training data, it transmits them to the AI ​​processing unit (145) in batches, thereby increasing the efficiency of retraining.

[0315] The training data generation unit (165) evaluates the performance of the retrained model when the retraining of the second artificial intelligence model (AI2) is completed. The training data generation unit (165) compares the 3D model generated by the retrained model with the actual final approved 3D model using test data and calculates a similarity score. The training data generation unit (165) calculates the similarity score using the Structural Similarity Index (SSIM) or the Mean Squared Error (MSE), and determines that the performance of the retrained model is excellent if the SSIM is 0.85 or higher or the MSE is below a preset threshold.

[0316] The training data generation unit (165) distributes the retrained model as the official model when the similarity score is above a preset threshold, and collects additional training data and repeats the retraining when the similarity score is below the threshold. When distributing the retrained model, the training data generation unit (165) backs up the previous model and replaces it with the new model, and monitors the model performance for a certain period (e.g., one week) after distribution so that if a problem occurs, it can roll back to the previous model.

[0317] The third AI model (AI3) learns past modification history data, final approved 3D virtual interior model data, and rental space type information to improve the accuracy of deriving optimal interior design patterns for each rental space type. The third AI model (AI3) continuously improves performance by learning modification history data from new projects, and over time, it can derive design patterns that more accurately reflect user preferences.

[0319] Although embodiments of the present invention have been described in more detail with reference to the attached drawings, the present invention is not necessarily limited to these embodiments and may be modified in various ways within the scope of the technical spirit of the present invention. Accordingly, the embodiments disclosed in the present invention are intended to explain, not limit, the technical spirit of the present invention, and the scope of the technical spirit of the present invention is not limited by these embodiments. Therefore, the embodiments described above should be understood as illustrative in all respects and not restrictive. The scope of protection of the present invention shall be interpreted by the claims below, and all technical spirits within an equivalent scope shall be interpreted as being included within the scope of rights of the present invention.

[0320] Therefore, other implementations, other embodiments, and equivalents to the claims also fall within the scope of the claims set forth below. Explanation of the symbols

[0321] 100: AI Model-Based Rental Space Matching Platform System 110: Matching Unit 120: Vendor Selection Unit 130: Design Input Unit 140: Model Generation Unit 150: Visualization Unit 160: Data Management Unit

Claims

Claim 1 A matching unit that receives rental space information from a building owner's terminal and provides information on eligible tenants and interior design companies; a tenant selection unit that selects tenants based on the rental space information transmitted from the matching unit, determines the priority of entry, and performs matching processing; a design input unit that receives interior design data from the representative terminal of the interior design company selected through the matching unit; a model generation unit that inputs the interior design data transmitted from the design input unit into a pre-trained artificial intelligence model to generate a 3D virtual interior model including the interior and exterior of the rental space; and a visualization unit that provides the 3D virtual interior model generated by the model generation unit to the building owner's terminal or the customer's terminal, receives a modification request, and transmits it to the model generation unit. and a data management unit that stores matching result data transmitted from the vendor selection unit, 3D virtual interior model data generated from the model generation unit, and modification history data transmitted from the visualization unit, and manages the entry history and interior progress status for each rental space; wherein the matching unit includes: an information input unit that receives rental space information including location information, area information, number of floors, floor height information, actual measurement dimension information, window location information, and rent information from the building owner's terminal, and converts the input rental space information into a structured data format; and a response providing unit that provides a list of target vendors transmitted from the vendor selection unit to the building owner's terminal and receives vendor selection information or an interior vendor request signal from the building owner's terminal.An interior design company database storing identification information, contact information, and specialized field information for each interior design company is constructed, and when an interior design company request signal is input through the response providing unit, an interior design company linkage unit generates a list of interior design companies corresponding to the leased space by linking with the constructed interior design company database, provides the generated list of interior design companies to the building owner's terminal, and receives the selected interior design company identification information and contact information from the building owner's terminal; and includes a data transmission unit that transmits structured rental space information converted from the information input unit to the company selection unit and the model generation unit, transmits company selection information entered through the response provision unit to the company selection unit, and transmits interior company identification information and contact information entered through the interior linkage unit to the design input unit; wherein the company selection unit receives structured rental space information transmitted from the data transmission unit, performs API communication with an external commercial area analysis system based on location information included in the received rental space information to query the number of floating people at the location, the number of surrounding competitors, and the distance to a public transportation station, assigns 20 points if the queried floating people number is 10,000 or more, 10 points if it is 5,000 or more but less than 10,000, and 0 points if it is less than 5,000, assigns 20 points if the number of surrounding competitors is less than 3, assigns 10 points if it is 3 or more but less than 5, and assigns 0 points if it is 5 or more, and assigns 20 points if the distance to a public transportation station is within 300m, A spatial analysis unit that assigns 10 points for distances exceeding 300m but within 500m and 0 points for distances exceeding 500m, calculates a location score by summing the assigned points and dividing by 60 to normalize them into a range of 0 to 100, and generates characteristic data for each rental space along with area information and rent information included in the rental space information;A tenant database is constructed to store industry information, average sales information, investment cost information, preferred location information, preferred area range information, history of past successful entry, and history of past failed entry by each company; a tenant lookup unit is configured to query the information stored in the constructed tenant database, and by comparing it with the characteristic data for each rental space transmitted from the spatial analysis unit, extract companies as eligible tenants whose location corresponds to a preferred location and whose area falls within the preferred area range, and generate tenant condition data for each company including the extracted industry information, average sales information, investment cost information, number of past entry attempts, and number of past successful entry; a tenant lookup unit is configured to combine the characteristic data for each rental space transmitted from the spatial analysis unit and the tenant condition data for each company transmitted from the tenant lookup unit, calculate a business feasibility score for each company using a pre-trained first artificial intelligence model, calculate a priority weight for entry based on the calculated business feasibility score, the location score transmitted from the spatial analysis unit, and the number of past entry attempts and successes transmitted from the tenant lookup unit, and according to the calculated priority weight A priority determination unit that sorts a list of prospective tenants; a matching execution unit that transmits the list of prospective tenants sorted by the priority determination unit to the response provision unit, receives tenant selection information transmitted from the data transmission unit, transmits matching request data including the rental space information, building owner contact information, and matching request time information to the headquarters system of the corresponding tenant, and transmits matching result data to the data management unit after the matching is completed;and a business data storage member that manages the aforementioned store database, and a business management member comprising a business evaluation member that calculates an entry success rate based on the past number of entry attempts and past number of successful entry attempts of each business transmitted from the business inquiry unit, wherein if the past number of entry attempts is 0, a basic entry success rate of 50 is assigned, and if the past number of entry attempts is 1 or more, the entry success rate is calculated as a percentage by multiplying the value obtained by dividing the past number of successful entry attempts by the past number of entry attempts by 100, and transmits the calculated entry success rate to the aforementioned priority determination unit;...including, and the above-mentioned store entry priority weight is calculated by the following [Equation 1], [Equation 1] W = α × B + β × L + γ × S (wherein W is the store entry priority weight, B is the business feasibility score calculated by the above-mentioned priority determination unit (normalized to a value between 0 and 100), L is the location score calculated by the above-mentioned spatial analysis unit (normalized to a value between 0 and 100), S is the store entry success rate calculated by the above-mentioned business evaluation unit (percentage, between 0 and 100), α, β, and γ represent the weighting coefficients for the business feasibility score, location score, and store entry success rate, respectively, and α, β, and γ are each values ​​between 0 and 1, α + β + γ = 1, and indicates that the larger the W value, the higher the store entry priority). The above-mentioned first artificial intelligence model learns past store entry success data, average sales data by industry, commercial area analysis data by region, and profitability data relative to rent, and the business feasibility score An AI model-based rental space matching platform system that improves the accuracy of calculation, wherein the priority determination unit calculates the expected monthly sales, expected investment recovery period, and expected operating profit margin for each company through the first AI model, and generates a business feasibility score normalized to a range of 0 to 100 by assigning higher values ​​as the expected monthly sales are higher, the expected investment recovery period is shorter, and the expected operating profit margin is higher, and wherein the priority determination unit excludes companies whose calculated entry priority weight is less than a preset minimum priority standard value from the list of eligible companies. Claim 2 delete Claim 3 In claim 1, the invention further comprises an industry database storing an average conversion rate, average transaction value, average cost ratio, standard monthly sales, standard payback period, and standard monthly net profit for each industry, including restaurants, cafes, unmanned stores, convenience stores, gyms, and academies; wherein the design input unit comprises: a communication connection unit that performs a communication connection with a representative terminal of an interior design company based on interior design company identification information and contact information transmitted from the data transmission unit; and a design data receiving unit that receives interior design drawing data including at least one of a 2D floor plan, an elevation view, and a cross-section view from the representative terminal of the interior design company through the communication connection unit, and extracts wall information, furniture placement information, finishing material information, and lighting information from the received interior design drawing data. and a data conversion unit that converts wall information, furniture placement information, finishing material information, and lighting information extracted from the design data receiving unit into a structured design data format and transmits the converted structured design data to the model generation unit; wherein the model generation unit receives structured design data transmitted from the data conversion unit and receives actual measurement dimension information, floor height information, and window location information of the rental space transmitted from the data transmission unit, generates a basic space model in the form of a three-dimensional rectangular prism having width, length, and height based on the received actual measurement dimension information and floor height information, and completes the three-dimensional basic space model by forming an opening at the corresponding location according to the received window location information; and an interior space partitioning model that generates an interior space partitioning model by placing walls that partition the interior space by applying wall information included in the structured design data transmitted from the data conversion unit to the three-dimensional basic space model generated from the space modeling unit, places furniture objects in three-dimensional coordinates by applying furniture placement information included in the structured design data to the generated interior space partitioning model, and for each placed furniture object, rotation and Internal modeling unit that generates a furniture placement model by performing resizing;An external modeling unit that applies external finishing material information, signboard information, and entrance information included in the structured design data transmitted from the data conversion unit to the exterior wall area of ​​the 3D spatial basic model generated from the spatial modeling unit, generates a signboard model based on the location, size, shape, and color of the signboard included in the signboard information, and generates an exterior model by placing the generated signboard model at a designated location on the exterior wall; a rendering unit that performs texture mapping according to the finishing material information included in the structured design data transmitted from the data conversion unit on the furniture placement model generated from the interior modeling unit and the exterior model generated from the external modeling unit, places a light source in a 3D space based on lighting type, lighting location, lighting brightness, and color temperature information according to the lighting information included in the structured design data, calculates the reflection, shadow, and contrast of light generated from the placed light source to perform photorealistic rendering to generate a final 3D virtual interior model, and generates image data captured from multiple viewpoints and 360-degree panoramic image data of the generated final 3D virtual interior model. AI processing unit comprising: a model storage member that stores a pre-trained second artificial intelligence model; and a model operation member that inputs structured design data transmitted from the data conversion unit into the second artificial intelligence model to generate 3D coordinate information, object placement information, and texture information, and transmits the generated 3D coordinate information, object placement information, and texture information to the spatial modeling unit, the internal modeling unit, the external modeling unit, and the rendering unit, respectively, wherein if there is information explicitly specified in the structured design data, such information is applied first, and if there is no explicit information, information generated by the second artificial intelligence model is used as auxiliary information;It includes, wherein the second artificial intelligence model is a generative artificial intelligence model that has learned past interior design data and 3D model data corresponding to the design, and is trained to receive a 2D design image as input and output 3D spatial coordinates, the location and size of an object, and material and color information; the rendering unit transmits the generated final 3D virtual interior model data, multiple viewpoint image data, and 360-degree panoramic image data to the visualization unit; the priority determination unit includes a commercial area data storage member that stores the number of floating population, the number of surrounding competitors, and distance information from public transportation stations included in the characteristic data for each rental space transmitted from the spatial analysis unit; a sales prediction member that queries the average conversion rate and average transaction value of the corresponding industry from the industry database based on industry information included in the entry condition data for each business transmitted from the business inquiry unit, calculates the expected number of customers by multiplying the number of floating population stored in the commercial area data storage member by the queried average conversion rate, and calculates the expected monthly sales by multiplying the calculated expected number of customers by the queried average transaction value; and the entry for each business transmitted from the business inquiry unit A profitability analysis unit that calculates monthly operating costs including labor costs, material costs, administrative costs, and marketing costs by multiplying the estimated monthly sales by the average cost ratio by industry retrieved from the industry database, based on investment cost information included in the condition data and the estimated monthly sales calculated from the sales forecasting unit; calculates monthly net profit by subtracting the monthly rent entered through the information input unit and the calculated monthly operating costs from the estimated monthly sales; and calculates the investment recovery period by dividing the initial investment cost included in the investment cost information by the monthly net profit.A score calculation member comprising: a score calculation member that combines the expected monthly sales calculated from the sales forecasting member, the investment recovery period calculated from the profitability analysis member, and the monthly net profit, calculates a business feasibility score according to the following [Mathematical Formula 2], adjusts the calculated business feasibility score to 100 if it exceeds 100 and to 0 if it is less than 0 to generate a final business feasibility score, and transmits the generated final business feasibility score to the priority determination member; [Mathematical Formula 2] B = δ × (R / R_ref) × 100 + ε × (P_ref / P) × 100 + ζ × (M / M_ref) × 100 (wherein B is the business feasibility score, R is the expected monthly sales calculated from the sales forecasting member, R_ref is the standard monthly sales of the corresponding industry stored in the industry database, P is the investment recovery period (in months) calculated from the profitability analysis member, P_ref is the standard investment recovery period (in months) of the corresponding industry stored in the industry database, and M is the A monthly net profit calculated from the profitability analysis unit, M_ref is the standard monthly net profit of the corresponding industry stored in the industry database, δ, ε, and ζ represent weighting coefficients for expected monthly sales, investment payback period, and monthly net profit, respectively, δ, ε, and ζ are values ​​between 0 and 1, δ + ε + ζ = 1, and the higher the final business feasibility score adjusted from the calculated B value to between 0 and 100, the higher the business feasibility; the priority determination unit calculates the entry priority weight W using the final business feasibility score transmitted from the score calculation unit as the input value of B in [Mathematical Formula 1], and the priority determination unit excludes companies from the list of eligible companies for entry for which the final business feasibility score transmitted from the score calculation unit is less than the preset minimum business feasibility standard score. This is an artificial intelligence model-based rental space matching platform system.

Images