Method, device, and system for providing solutions for corporate grouping and intra-group network construction through ai-based corporate data analysis

Abstract

One embodiment of the present invention relates to a method, apparatus, and system for providing a solution for corporate grouping and network construction within a group through artificial intelligence-based corporate data analysis, which collects and groups corporate data, analyzes suppliers and customers within the group in a linked manner, or provides a list of companies and movement paths based on search conditions.

Description

Technology Field

[0001] The following embodiments relate to a method, apparatus, and system for providing a solution for corporate grouping and network construction within a group through AI-based corporate data analysis, which collects and groups corporate data, analyzes suppliers and customers within the group in a linked manner, or provides a list of companies and movement paths based on search conditions. Background Technology

[0002] Conventional corporate data analysis has primarily relied on financial statements, corporate registration information, and structured data obtained through a single API. While this approach can be utilized to classify companies by industry or size, it has limitations in reflecting substantive network characteristics, such as actual purchasing and sales relationships or transaction history.

[0003] Furthermore, unstructured document data, such as tax invoices and contracts serving as proof of transaction history, could not be effectively utilized in existing systems. While advancements in OCR technology enabled the extraction of some data, there was a lack of functionality to automatically verify the reliability of the extracted data and reinforce it by linking it with corporate information databases.

[0004] Furthermore, existing systems did not sufficiently provide functions such as analyzing supplier-sales relationships between companies to identify potential new business partners, searching for companies based on user-entered conditions, or optimizing delivery routes. Prior art literature

[0005] (Patent Document 0001) KR 10-2720676 B (Patent Document 0002) KR 10-2420626 B (Patent Document 0003) KR 10-2809718 B The problem to be solved

[0006] The problem that an embodiment of the present invention aims to solve is to provide a method, apparatus, and system for providing a solution for corporate grouping and network construction within a group through AI-based corporate data analysis, which overcomes the limitations of conventional corporate data analysis and business partner matching technologies as described above by collecting corporate data from user terminals and corporate information APIs, performing data verification through OCR-based unstructured document recognition and reliability scores to accurately secure corporate information, analyzing corporate data to accurately classify companies into multiple groups based on industry, products, and trade items to match new business partners, searching and exploring corporate information by reflecting search conditions entered from a user terminal, and analyzing and providing the delivery location routes of related companies, thereby supporting the discovery of inter-company transaction opportunities and the optimization of logistics routes. means of solving the problem

[0007] According to one embodiment, a device includes a processor, memory, a communication module, and a non-transient storage medium, and a method for providing a solution for corporate grouping and network construction within a group through AI-based corporate data analysis, wherein the method is performed by the processor by executing a program stored in the non-transient storage medium, the method comprises: a step of collecting corporate data, which is data of a first company, from a user terminal and a corporate information API; a step of inputting the corporate data into an AI model to classify the first companies into a plurality of first groups; a step of mutually analyzing suppliers and customers for each first group to select new business partner candidates, including new suppliers or new customers, for each first company; a step of transmitting information on the new business partner candidates to the user terminal of the first company for which the new business partner candidates have been selected; and a step of transmitting information on the first company corresponding thereto to the user terminal of the new business partner candidates.

[0008] Additionally, the method further comprises the steps of: receiving search conditions from the user terminal; extracting a second company, which is the first company corresponding to the search conditions, among the first companies; calculating a movement path between delivery locations of the second companies based on the information of the second company; and transmitting the information of the second company and the movement path between delivery locations to the user terminal; and the step of collecting company data comprises: receiving a tax invoice, a transaction statement, or a contract from the user terminal; extracting a transaction industry, transaction items, and a transaction amount from the tax invoice, transaction statement, or contract based on an OCR module and designating them as first user data; receiving transaction history information including sales source information, purchasing source information, transaction industry, transaction items, and a transaction amount from the user terminal and designating it as second user data; comparing the first user data and the second user data to calculate a reliability score based on the content match rate; and if the reliability score is below a predetermined threshold, transmitting an information re-examination request message to the user terminal. The method may include: a step of adding the first user data and the second user data to the corporate data when the above reliability score is greater than or equal to a predetermined threshold; and a step of receiving credit rating information and corporate registration information corresponding to each of the first companies from the corporate information API and adding them to the corporate data.

[0009] And, the step of classifying first companies into multiple first groups based on the above company data may include: a preprocessing step of converting industry names, product names, and trade item names included in the above company data into standardized keywords through a natural language processing process including morphological analysis, stop word removal, and synonym dictionary mapping; a step of generating a first vector by vectorizing the standardized keywords extracted for each first company; and a step of clustering the first companies into first groups based on the first vectors.

[0010] In addition, the step of selecting the new business partner candidates comprises: a step of calculating a first centrality index for each of the 3-1 companies included in the first group, based on transaction amount, transaction frequency, number of sales partners, number of purchase partners, and credit rating information for each of the 1st groups; a step of designating a 3-1 company whose first centrality index is greater than or equal to a predetermined threshold as a 3-2 company; a step of designating each of the 3-2 companies as a central node and designating a 1 company that has a transaction history with the 3-2 company as a peripheral node to generate a first network graph for each 3-2 company; a step of calculating an edge weight corresponding to each edge based on transaction amount, transaction frequency, and transaction item similarity for the first network graph; a step of designating a peripheral node directly connected to only one of the peripheral nodes, either the central node or another peripheral node, as a terminal node; a step of tracing a first path, which is the shortest path connecting from the central node to each terminal node, and summing the edge weights of the edges included in each first path to calculate a first network length for each terminal node; For the first network graph above, a step of calculating the first network average length, the first network maximum length, the first network minimum length, and the first network length standard deviation based on the first network length; a step of designating a neighbor node connected to three or more other neighbor nodes or connected to a 'central node and two or more other neighbor nodes' as an overlapping node; a step of designating, for each overlapping node, a 'neighbor node among the other connected neighbor nodes that is not included in the first path of the overlapping node' as a radiating node; a step of tracing a second path, which is the shortest path connecting from the radiating node to the terminal node, and calculating the second network length for each radiating node by summing the edge weights of the edges included in each second path;A step of calculating a second network average length, a second network maximum length, a second network minimum length, and a second network length standard deviation based on the second network length for the first network graph; a step of calculating a length distribution homogeneity index, a bottleneck index, and a distribution bias index based on the first network average length, the first network maximum length, the first network minimum length, the first network length standard deviation, the second network average length, the second network maximum length, the second network minimum length, the second network length standard deviation, and the number of radiating nodes for the first network graph; a structural determination step of determining the structural properties of the first network graph as one of compact homogeneous type, bottleneck dominant type, heterogeneous risk type, and bias type according to pre-set standard conditions for the network characteristic index; a step of extracting a third-third company, which is a third-second company that matches the first company among third-second companies, based on a pre-specified preference structure for the first company; and a step of designating the first network graph of the third-third company as the second network graph. and may include the step of designating the first list of companies included in the second network graph as new business partner candidates.;

[0011] In addition, the step of calculating the first centrality index comprises: a step of log-normalizing the total sum of transaction amounts for each of the 3-1 companies; a step of standardizing the transaction frequency for each of the 3-1 companies by dividing it by the average transaction frequency of the corresponding 1 group; a step of converting the number of sales outlets and the number of purchase outlets for each of the 3-1 companies into ratios relative to the maximum value within the corresponding 1 group, respectively; a step of converting credit rating information for each of the 3-1 companies into a credit index pre-specified by grade; a step of calculating the second centrality index by linearly summing the total sum of transaction amounts, transaction frequency, number of sales outlets, number of purchase outlets, and credit index according to pre-specified weights for each of the 3-1 companies included in the 1 group; and a step of calculating the third centrality index by normalizing the second centrality index calculated for each of the 3-1 companies included in the 1 group according to the interval between the minimum and maximum values ​​of all second centrality indices calculated for the corresponding 1 group. and a step of designating the third centrality index as the first centrality index corresponding to the 3-1 enterprise; wherein the step of calculating the network characteristic indicator comprises: a step of calculating a first coefficient of variation by dividing the first network average length by the first network length standard deviation; a step of calculating a second coefficient of variation by dividing the second network average length by the second network length standard deviation; a step of calculating a length distribution homogeneity index by weighted averaging the first coefficient of variation and the second coefficient of variation; a step of calculating the number of radial connections per radial node, which is the number of edges connected to each of the radial nodes; a step of calculating a first radial concentration by dividing the number of radial connections per radial node by the average value of the number of radial connections per radial node; a step of calculating a second radial concentration, which is 'the number of radial nodes where the first radial concentration is greater than or equal to a predetermined threshold'; and a step of calculating a bottleneck indicator, which is the value obtained by dividing the second radial concentration by the number of surrounding nodes.The method may include: a step of calculating a first bias coefficient of '(maximum length of the first network + minimum length of the first network - 2 × average length of the first network) / (maximum length of the first network - minimum length of the first network)'; a step of calculating a second bias coefficient of '(maximum length of the second network + minimum length of the second network - 2 × average length of the second network) / (maximum length of the second network - minimum length of the second network)'; a step of calculating a first bias weight of 'number of terminal nodes / (number of terminal nodes + number of radiating nodes)'; a step of calculating a second bias weight of 'number of radiating nodes / (number of terminal nodes + number of radiating nodes)'; and a step of calculating a bias index of the distribution of 'first bias coefficient × first bias weight + second bias coefficient × second bias weight'.

[0012] A device according to one embodiment may be combined with hardware and controlled by a computer program stored on a medium to execute the method of any one of the methods described above. Effects of the invention

[0013] According to one embodiment, through the process of collecting and grouping corporate data, companies can be efficiently clustered based on actual transaction networks, going beyond simple industry classification.

[0014] In addition, by conducting an analysis of the interconnectedness between suppliers and customers, it is possible to identify potential new suppliers and customers beyond existing business relationships, thereby expanding business-to-business transaction opportunities.

[0015] In addition, by receiving search conditions from the user terminal and calculating and providing the delivery route to the company that meets the conditions, an optimized route can be secured in logistics and supply chain management.

[0016] Furthermore, by standardizing and grouping the industries, products, and trade items included in corporate data based on natural language processing, misclassification caused by synonyms or similar concepts is reduced, and more sophisticated clustering is possible.

[0017] In addition, by calculating centrality indices and performing structural analysis based on network graphs, the structural properties of inter-firm transaction networks (compact / homogeneous, bottleneck-dominant, heterogeneous risk, and biased) can be determined, thereby improving corporate risk management and the reliability of new business partner recommendations.

[0018] In addition, by comprehensively reflecting network characteristic indicators (length distribution homogeneity index, bottleneck indicator, distribution bias index), complex stability can be evaluated without relying on a single indicator.

[0019] In addition, since it includes a reliability score-based data verification procedure, it can correct discrepancies between OCR and user input data to ensure the accuracy of corporate data. Brief explanation of the drawing

[0020] FIG. 1 is a schematic diagram showing a system for providing solutions for corporate grouping and network construction within a group through artificial intelligence-based corporate data analysis according to an embodiment of the present invention. FIG. 2 is a flowchart illustrating a method for providing a solution for corporate grouping and network construction within a group through artificial intelligence-based corporate data analysis according to an embodiment of the present invention. FIG. 3 is a flowchart illustrating a method for providing a solution for corporate grouping and network construction within a group through artificial intelligence-based corporate data analysis according to another embodiment of the present invention. FIG. 4 is a flowchart illustrating the step of collecting corporate data in a method for providing a solution for corporate grouping and network construction within a group through artificial intelligence-based corporate data analysis according to an embodiment of the present invention. FIG. 5 is a flowchart illustrating the step of classifying first companies into a plurality of first groups in a method for providing a solution for corporate grouping and network construction within a group through artificial intelligence-based corporate data analysis according to an embodiment of the present invention. Specific details for implementing the invention

[0021] 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.

[0022] 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.

[0023] 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.

[0024] When it is stated that a component is "connected" to another component, it should be understood that it may be directly connected to or joined to that other component, or that there may be other components in between.

[0025] 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.

[0026] In particular, where a 'step' in this specification is described as 'comprising' one or more detailed steps or sub-steps, said 'step' may be interpreted as including its own basic processing step while simultaneously performing the described detailed steps as well.

[0027] For example, if it is stated that 'a step of doing B to A' includes 'a step of doing D to C; a step of doing F to E; and a step of doing H to G,' the 'step of doing B to A' may be interpreted not merely as the basic operation of doing B to A, but as a configuration that performs detailed procedures together, such as a step of doing D to C, a step of doing F to E, and a step of doing H to G.

[0028] Accordingly, the above configuration does not exclude various sub-procedures included within the scope of execution of the corresponding step, and may be included within the scope of the present invention even if other procedures or means performing substantially the same or equivalent functions are substituted.

[0029] Expressions such as 'end part', 'both ends', 'one end', 'other end', and 'side end' of a component can be interpreted as referring to at least / any one of the end parts of that component.

[0030] In the description of the present invention, 'a method in which a device comprises a processor, a memory, a communication module, and a non-transient storage medium, and a program stored in the non-transient storage medium is executed by the processor,' the term 'method' may be interpreted as referring to the program stored in the non-transient storage medium itself or a part of the program.

[0031] The term 'Return' as used in the description of the present invention may refer to a result value being output, returned, or returned from a method, procedure, function, etc. used in a given program language / structure.

[0032] 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. For example, the term 'artificial intelligence model' may be selected from one or more of known general artificial intelligence models.

[0033] 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.

[0034] According to one embodiment, a device includes a processor, memory, a communication module, and a non-transient storage medium, and a method for providing a solution for corporate grouping and network construction within a group through AI-based corporate data analysis, wherein the method is performed by the processor by executing a program stored in the non-transient storage medium, the method comprises: a step of collecting corporate data, which is data of a first company, from a user terminal and a corporate information API; a step of inputting the corporate data into an AI model to classify the first companies into a plurality of first groups; a step of mutually analyzing suppliers and customers for each first group to select new business partner candidates, including new suppliers or new customers, for each first company; a step of transmitting information on the new business partner candidates to the user terminal of the first company for which the new business partner candidates have been selected; and a step of transmitting information on the first company corresponding thereto to the user terminal of the new business partner candidates.

[0035] The step of collecting corporate data, which is the data of the first company, from user terminals and corporate information APIs is a step of acquiring structured and unstructured data regarding the first company (including the company that owns or manages the user terminal and companies linked as sales or purchasing partners of said company) in a multi-dimensional manner and normalizing it into an integrated schema that can be utilized for subsequent analysis.

[0036] User terminals may include PCs, tablets, or mobile terminals linked to accounting, purchasing, or sales systems, and transmitted data may include original files and metadata such as tax invoices, transaction statements, contracts, incoming and outgoing inventory records, item code tables, and delivery address books.

[0037] The corporate information API may consist of credible corporate credit rating agencies, commercial registration and business registration databases, and Standard Industrial Classification APIs, and provides items such as corporate name, business registration number, industry code, head office and branch addresses, credit rating, and rating assignment date.

[0038] In the data collection phase, OCR preprocessing is performed on unstructured documents received from user terminals to extract transaction industry names, transaction item names, unit prices, quantities, supply amounts, tax amounts, transaction dates, and counterparty identifiers (business registration numbers, corporate names, etc.). After checking consistency at the record level through key matching (corporate identifiers, address alignment, name similarity, etc.) with corporate information API responses, duplicate transactions (identical document numbers, amounts, dates), split transactions (multiple tax invoices for the same order), and reverse journal entry transactions (returns, deductions) are identified using standard rules and loaded into an integrated table.

[0039] Here, in cases where names differ but are determined to be the same company (e.g., “Ganada Co., Ltd.” and “Ganada (Co., Ltd.)”), they can be integrated into the same identifier by utilizing predefined string normalization rules and the matching of the representative name and industry code from the corporate information API. Additionally, to ensure data reliability, the matching rate between transaction items and amounts extracted from the user terminal and past reporting patterns based on the corporate information API is calculated; if the rate falls below a threshold, the corresponding record can be marked as pending to be excluded from subsequent analysis, or a notification requesting a review can be sent to the user terminal.

[0040] The step of inputting the above-mentioned corporate data into an artificial intelligence model to classify the first companies into multiple first groups is a step of converting the collected and normalized corporate data into a vector representation through natural language processing and numerical transformation, and forming clusters by calculating the similarity between companies based on this.

[0041] In the classification stage, a standardized keyword list is generated by performing morphological analysis, stop word removal, and synonym / superordinate dictionary mapping on industry names, product names, and trade item names, and a first-firm keyword vector is generated using keyword frequency, TF-IDF, or pre-trained embeddings.

[0042] At the same time, numerical indicators such as total transaction amount, transaction frequency, average unit price fluctuation rate, major suppliers, and the proportion of major sales sources can be standardized and combined into an auxiliary feature vector.

[0043] Using the first vectors generated in this way as input, data are classified / clustered based on conventional artificial intelligence models, such as K-means, DBSCAN, or hybrid clustering (hierarchical + density-based) models.

[0044] For example, if companies A, B, and C, which have high keyword weights related to “electronic component assembly and wholesale / retail,” and companies D, E, and F, which have high keyword weights related to “food raw material supply and processing,” form different centers, the former are divided into the electronic components cluster and the latter into the food processing cluster. As a result of this step, each first-tier company is fundamentally assigned to exactly one first-tier group (classification result), and this is subsequently used as the standard group for candidate recommendations in the network analysis stage.

[0045] The step of selecting new business partner candidates, including new suppliers or new sales partners, for each first company by mutually analyzing suppliers and sales partners for each of the first groups is a step of estimating counterparty companies with high transaction potential by combining the actual transaction network and the item similarity network within the group.

[0046] The step of selecting new business partner candidates first involves constructing a transaction history-based network graph within Group 1. Nodes represent the first company, and edges represent the transaction relationship between the two companies; edge weights can utilize a composite score calculated from transaction amount, transaction frequency, and item similarity (such as standardized keyword cosine similarity).

[0047] The step of transmitting new business partner candidate information to the user terminal of the first company, where the above-mentioned new business partner candidate has been selected, is a step of supporting the user in making a decision by providing explanatory information including the reasons for candidate selection and expected effects along with the list of selected candidates.

[0048] The transmitted information may include additional data generated based on a separate AI model, such as the candidate company's basic identification information (corporate name, business registration number, industry code), a list of items with high expected synergy (representative item name, standard code, predicted unit price range), estimated transaction potential (expected monthly transaction amount and frequency), credit summary (rating and date of assignment), risk warnings (recent arrears, rating downgrade, etc.), and recommendation priority. Furthermore, to enable users to take prompt follow-up actions, if contact information for a representative exists, the corresponding metadata (main email and phone number) or identifiers for integration with the internal CRM may also be provided. If necessary, users can provide feedback to the system by classifying candidates as "Interested," "Pending," or "Excluded," and this feedback may be reflected in subsequent parameter updates for the recommendation model.

[0049] The step of transmitting information of the corresponding first company to the user terminal of the aforementioned new client candidate is a step of providing customized proposal information to the candidate company side as well in order to complete the mutual recommendation structure.

[0050] At this stage, a summary of the primary counterparty's basic information, estimated required items and order volume, recent order cycle, desired delivery conditions, and payment terms is transmitted together so that the candidate company can consider it as a new revenue opportunity.

[0051] For example, a proposal may be provided to a user terminal of Company C in the form of “Company A: Monthly average predicted demand for capacitor C-1206 specification X thousand units, similar procurement pattern within the last 6 months, desired delivery time 7 days, preferred payment terms 30-day credit”.

[0052] Information transmission is carried out via an encrypted channel in accordance with mutual consent and personal information and trade secret protection policies, and the receiving user may request a counter-proposal or consultation schedule through an “interest response.”

[0053] When mutual interest is matched, the system can pre-calculate and provide basic route information based on the delivery location addresses of both parties. At this time, a basic route is extracted considering address normalization (road name address and coordinate conversion) and trunk weighting (distance, estimated transport time, and tolls); additionally, if there is an option to reflect traffic congestion by time of day specified by the user, alternative routes applying those conditions can also be presented.

[0054] Additionally, the method further comprises the steps of: receiving search conditions from the user terminal; extracting a second company, which is the first company corresponding to the search conditions, among the first companies; calculating a movement path between delivery locations of the second companies based on the information of the second company; and transmitting the information of the second company and the movement path between delivery locations to the user terminal; and the step of collecting company data comprises: receiving a tax invoice, a transaction statement, or a contract from the user terminal; extracting a transaction industry, transaction items, and a transaction amount from the tax invoice, transaction statement, or contract based on an OCR module and designating them as first user data; receiving transaction history information including sales source information, purchasing source information, transaction industry, transaction items, and a transaction amount from the user terminal and designating it as second user data; comparing the first user data and the second user data to calculate a reliability score based on the content match rate; and if the reliability score is below a predetermined threshold, transmitting an information re-examination request message to the user terminal. The method may include: a step of adding the first user data and the second user data to the corporate data when the above reliability score is greater than or equal to a predetermined threshold; and a step of receiving credit rating information and corporate registration information corresponding to each of the first companies from the corporate information API and adding them to the corporate data.

[0055] The step of receiving search conditions from the user terminal is a step of receiving filter values ​​representing the attributes, transaction preferences, and logistics constraints of the company the user intends to search for via a native UI or web screen and converting them into a standardized query structure.

[0056] Search criteria may include industry code, name of major products or procurement items, minimum and maximum transaction amounts, minimum credit rating, preferred payment terms, available delivery days and times, available delivery zone (city / province, radius in kilometers), vehicle tonnage, loading method (whether a forklift is required), etc.

[0057] In this step, keywords entered in natural language (e.g., “HMR packaging company capable of delivering twice a week in Dongjak-gu, Seoul”) are mapped to standard fields through morphological analysis and synonym dictionary mapping, missing fields are corrected to default values, and overly narrow combinations are encouraged to re-enter by suggesting a recommended range.

[0058] Among the above first companies, the step of extracting the second company, which is the first company corresponding to the search condition, is a step of selecting a candidate group by executing the standardized query of the previous step in the integrated company database and calculating a ranking by multi-criteria scoring.

[0059] First, only companies that satisfy essential filters such as industry, product, credit rating, and region are selected, and second, suitability based on transaction history (past processing volume of the relevant item, on-time delivery rate, return rate), capacity suitability (daily throughput margin, warehouse available area), and policy suitability (desired payment conditions, minimum transaction quantity) are scored using item-specific weights.

[0060] In this stage, weights can be personalized by reflecting user preferences already held by the system (e.g., excluding company types previously marked as 'pending'), and in the event of a tie, the reliability of recent data and the latest update date are given priority. For example, regarding the conditions “refrigerated food packaging, 20-kilometer radius from Seoul, credit rating A or higher, 3 deliveries per week,” companies with verified refrigeration capacity and available slots for each day of the week may be ranked higher.

[0061] Based on the information of the second company mentioned above, the step of calculating the movement path between the delivery locations of the second companies is to normalize the delivery address of each extracted candidate company and the reference location designated by the user (outbound warehouse, factory, logistics center, final delivery destination) into coordinates, and then optimize the order of visits to calculate the expected travel route and required time.

[0062] In this step, addresses are normalized into a road name, lot number, and coordinate system, and pathfinding is performed by setting edge costs in the road network graph as distance or time. When there are multiple delivery locations, nearest insertion, 2-exchange, and time window constraint heuristics are applied to approximate the vehicle path problem, and user constraints (avoiding lunch hours, nighttime prohibited areas, and roads restricted to cargo trucks) are reflected as options. Additionally, a realistic total time is calculated by including the ride-sharing effect (reduction in empty vehicle rate) when combining multiple candidates within the same area, estimated picking and loading times, and penalties for bypassing areas where stopping is not possible.

[0063] For example, a comparison result can be generated such that the visit sequence of Gangseo-gu → Guro-gu → Dongjak-gu is shortened by 15 minutes compared to Gangseo-gu → Dongjak-gu → Guro-gu, and increases by 5 minutes when truck-restricted road avoidance is applied.

[0064] The step of transmitting the information of the second company and the route between the delivery locations to the user terminal involves providing a package of the optimal route, alternative route, estimated transport timetable, and recommended departure time, along with a key summary of the selected candidate company. The transmitted information may include basic information for each candidate company, reasons for recommendation (basis for item suitability, capacity suitability, and credit suitability), estimated transaction potential, and constraint warnings (time window conflict, vehicle tonnage mismatch), as well as route coordinate sequences for map-based visualization, estimated arrival and departure times for each waypoint, and total driving distance and fuel cost estimates. The user can recalculate the route through interaction, such as excluding a specific candidate or fixing it as a priority destination on the interface, and the recalculation results are displayed in the form of a comparison table on the same screen.

[0065] The aforementioned company search / exploration processes can be processed / performed separately from / in parallel with the business partner matching process, as illustrated in FIG. 3. In other words, a platform according to an embodiment of the present invention may include both a business partner matching function and a search function.

[0069] After the step of calculating the movement path between the delivery locations mentioned above, the method may further include a step of automatically listing component suppliers and sub-process partner companies required for the production of the finished product based on the items and process flow of the finished product that the user intends to manufacture.

[0070] In this stage, clustering results and transaction history data stored in the corporate database are utilized to map partner companies linked to the finished product production flow by process, and the results are provided to the user terminal.

[0071] Furthermore, it may include a step of optimizing the entire supply chain route by combining location information, transportation distance, estimated lead time, and cost data of the aforementioned listed partner companies.

[0072] In this step, the supply chain route is simulated based on the final delivery location, and a site layout plan suitable for the supply chain flow is automatically calculated based on the results. One or more generated layout scenarios are visualized and provided to the user terminal, allowing the user to expect benefits such as reduced production time, lower transportation costs, and improved delivery responsiveness.

[0073] When companies located within the same smart building cooperate in the production of finished products, the present invention can automatically control the entire supply chain process in real time by integrally controlling truck entry control, cargo elevator scheduling, and warehouse loading and dispatch processes.

[0074] For example, when logistics cooperation is required between tenant companies within the same building or complex, truck entry control, freight elevator calling, and warehouse loading and dispatch scheduling are automatically controlled, enabling the entire supply chain process to be automated in real time.

[0075] One embodiment may further include a step of automatically performing escrow payment or point / coin-based step-by-step payment and settlement depending on the contract conditions and the progress of the process stage.

[0076] Based on contract terms and the progress of each construction phase, decisions regarding payment suspension or refunds, as well as phased payments, can be automatically made based on the completion status of each process, inspection results, user terminal input, sensor measurements, and CCTV object recognition. This automated decision-making function enhances contract stability and transaction transparency.

[0083] The step of receiving a tax invoice, transaction statement, or contract from the user terminal involves uploading the original evidence through a secure channel, verifying the document metadata and file integrity, and loading it into a collection queue. The document type, issue date, document number, supplier / recipient identifier, total amount, and whether attachments are attached are recorded as primary metadata, and a file hash is stored to prevent duplicate uploads. In the case of batch uploading multiple files, transmission failure items are separated and sent to a retransmission queue to protect the user experience.

[0084] The step of extracting the transaction industry, transaction items, and transaction amounts from the aforementioned tax invoice, transaction statement, or contract based on the OCR module and designating them as first user data is a step of generating structured data by combining document layout recognition and field extraction rules. It performs table header candidate detection, cell boundary restoration, and item line parsing, and normalizes typos and abbreviations by matching item name notations with a standard material code dictionary. The amount field is verified for consistency using currency unit and tax separation rules, and a pending flag is assigned if the total amount and line sum are inconsistent. The extraction results are recorded in the first user data in the form of “Document Unique Number-Line Index-Item Code-Quantity-Unit Price-Supply Amount-Tax Amount-Transaction Date”.

[0085] The step of receiving transaction history information, including sales source information, purchasing source information, transaction industry, transaction items, and transaction amounts, from the above user terminal and designating it as second user data involves loading CSV and API responses exported from the ERP and accounting systems into a standard schema and verifying essential keys (customer identifier, voucher date, amount). The customer name is normalized and matched with the standard business name of the enterprise information API to eliminate duplicates, and the item name is mapped to the standard code from the previous step to align item standards between different sources. At this time, unit conversion (box to individual item) is applied to unify the standard units of amount and quantity.

[0086] The step of calculating a reliability score based on the content matching rate by comparing the above-mentioned first user data and second user data is a step of quantifying accuracy through cross-verification at the same period, same business partner, and same item level. It calculates the item matching rate, amount matching rate, tax amount matching rate, date proximity rate, and document number matching rate, each assigned a weight, and deducts points by applying outlier detection rules (e.g., exceeding the allowable range for amount errors, omission of reverse journal entries).

[0087] For example, a high reliability score is assigned if the amount and tax amount match exactly, the item name is identical in standard code, and the date is within 3 business days, while points are deducted if the document number is missing or the item code fails to match. The final score is normalized to a value between 0 and 1 and used as a branching criterion for subsequent processing.

[0088] The step of sending an information review request message to the user terminal when the above reliability score is below a pre-specified threshold is a step of providing feedback including discrepancies and correction guides in real time. The message may include a list of discrepancies in records, expected corrections (e.g., item name candidates, document number format suggestions), recommended actions (re-upload, creation of missing slips), and a toggle to select whether to apply automatic correction. If automatic correction is allowed, a re-verification loop is performed by applying proposed rules within a safe range.

[0089] If the above reliability score is above a pre-specified threshold, the step of adding the first and second user data to the corporate data involves permanently reflecting the verified data in the ledger table and creating an index so that it can be immediately used for future analysis. This involves constructing a composite index of the business partner, item, and period, an index of the corporate identifier, etc., and incrementally updating the aggregation table (monthly transaction amount, proportion by item) for the recently added data range. This ensures the input quality for the clustering and recommendation operations in the next step.

[0090] The step of receiving credit rating information and corporate registration information corresponding to each of the first companies from the above-mentioned corporate information API and adding them to corporate data is a step of periodically synchronizing credit ratings, rating assignment dates, debt ratio ranges, business suspension / closure status, and history of changes in business status from external credible data sources and reflecting them in risk management.

[0091] In this stage, timestamps of API responses are stored to enable timeliness verification. If a credit rating downgrade event is detected, the priority of recommended candidates, including the company in question, can be automatically adjusted, or a lower threshold can be applied to exclude them from the candidate pool. Additionally, discrepancies between the registered address and the actual delivery address can be detected and displayed as a warning when calculating logistics routes.

[0092] And, the step of classifying first companies into multiple first groups based on the above company data may include: a preprocessing step of converting industry names, product names, and trade item names included in the above company data into standardized keywords through a natural language processing process including morphological analysis, stop word removal, and synonym dictionary mapping; a step of generating a first vector by vectorizing the standardized keywords extracted for each first company; and a step of clustering the first companies into first groups based on the first vectors.

[0093] The step of classifying the first companies into multiple first groups based on the aforementioned corporate data involves converting the structured and unstructured text attributes possessed by each company into standardized vector representations, and then performing clustering based on similarity to form groups of similar industries and product categories. This process is broadly divided into a preprocessing step, a vectorization step, and a clustering step.

[0094] The preprocessing stage involves converting industry names, product names, and trade item names included in the corporate data into normalized keywords using natural language processing techniques. In this stage, a morphological analyzer is used to separate sentences into root units such as nouns, verbs, and adjectives, and meaningless particles, conjunctions, and general stop words (e.g., “and,” “and,” “related”) are removed from the analysis results. Additionally, a thesaurus is utilized to standardize words with identical meanings in order to reduce the variance in expressions for industries and products.

[0095] For example, “signboard,” “information board,” and “signpost” can all be standardized into “signboard,” and “food manufacturing” and “food processing” can be integrated into “food manufacturing.” Keywords processed in this way are designated as a standardized set of keywords representing the core attributes of each company.

[0096] The step of generating a first vector by vectorizing the standardized keywords extracted for each of the first companies is a step of projecting the keyword set, which is the result of preprocessing, into a vector space capable of numerical computation.

[0097] At this time, TF-IDF-based weights can be assigned to each keyword, or multidimensional vectors can be produced using pre-trained language models such as Word2Vec, FastText, and BERT embeddings.

[0098] For example, if Company A's keywords are {"signboard," "aluminum," "manufacturing"} and Company B's keywords are {"information board," "steel plate," "installation"}, "information board" is unified to "signboard" during the preprocessing stage, so the two companies generate a vector with some common components. This first vector numerically represents the characteristics of the companies' business sectors and products.

[0099] The step of clustering the first companies into a first group based on the first vectors above is a step of calculating the similarity between companies to form multiple groups of companies.

[0100] In this step, distances or angles between vectors are calculated using metrics such as cosine similarity and Euclidean distance, and algorithms such as K-means, DBSCAN, and Hierarchical Clustering can be applied.

[0101] For example, companies with similar industry, product, or item keywords are grouped together, and “metal sign manufacturing companies,” “plastic sign manufacturing companies,” and “LED sign manufacturing companies” can be classified into a single group. These generated groups are subsequently used as a basis for recommending potential new clients or analyzing network structures.

[0102] In addition, the step of selecting the new business partner candidates comprises: a step of calculating a first centrality index for each of the 3-1 companies included in the first group, based on transaction amount, transaction frequency, number of sales partners, number of purchase partners, and credit rating information for each of the 1st groups; a step of designating a 3-1 company whose first centrality index is greater than or equal to a predetermined threshold as a 3-2 company; a step of designating each of the 3-2 companies as a central node and designating a 1 company that has a transaction history with the 3-2 company as a peripheral node to generate a first network graph for each 3-2 company; a step of calculating an edge weight corresponding to each edge based on transaction amount, transaction frequency, and transaction item similarity for the first network graph; a step of designating a peripheral node directly connected to only one of the peripheral nodes, either the central node or another peripheral node, as a terminal node; a step of tracing a first path, which is the shortest path connecting from the central node to each terminal node, and summing the edge weights of the edges included in each first path to calculate a first network length for each terminal node; For the first network graph above, a step of calculating the first network average length, the first network maximum length, the first network minimum length, and the first network length standard deviation based on the first network length; a step of designating a neighbor node connected to three or more other neighbor nodes or connected to a 'central node and two or more other neighbor nodes' as an overlapping node; a step of designating, for each overlapping node, a 'neighbor node among the other connected neighbor nodes that is not included in the first path of the overlapping node' as a radiating node; a step of tracing a second path, which is the shortest path connecting from the radiating node to the terminal node, and calculating the second network length for each radiating node by summing the edge weights of the edges included in each second path;A step of calculating a second network average length, a second network maximum length, a second network minimum length, and a second network length standard deviation based on the second network length for the first network graph; a step of calculating a length distribution homogeneity index, a bottleneck index, and a distribution bias index based on the first network average length, the first network maximum length, the first network minimum length, the first network length standard deviation, the second network average length, the second network maximum length, the second network minimum length, the second network length standard deviation, and the number of radiating nodes for the first network graph; a structural determination step of determining the structural properties of the first network graph as one of compact homogeneous type, bottleneck dominant type, heterogeneous risk type, and bias type according to pre-set standard conditions for the network characteristic index; a step of extracting a third-third company, which is a third-second company that matches the first company among third-second companies, based on a pre-specified preference structure for the first company; and a step of designating the first network graph of the third-third company as the second network graph. and may include the step of designating the first list of companies included in the second network graph as new business partner candidates.;

[0103] The step of calculating the first centrality index for each of the 3-1 companies included in the first group, based on transaction amount, transaction frequency, number of sales outlets, number of purchase outlets, and credit rating information for each of the above-mentioned first groups, is a process of quantifying the importance of individual companies within the transaction network.

[0104] For example, a company with large transaction amounts, high transaction frequency, and a diverse range of sales and purchasing sources can be considered to have significant influence within the network. Additionally, since companies with superior credit ratings exhibit higher transaction stability, this can be weighted and reflected when calculating the centrality index.

[0105] The step of designating the 3-1 company, whose 1st centrality index is greater than or equal to a previously specified threshold, as the 3-2 company is a process of selecting core node candidates in network analysis.

[0106] The threshold can be set, for example, to a value greater than the mean centrality index within the group, and companies exceeding this are considered to play a key role in the transaction structure within the group.

[0107] The step of designating each of the aforementioned 3-2 companies as a central node and designating the 1 company, which has a transaction history with the 3-2 companies, as a peripheral node to generate the 1st network graph for each 3-2 company is a process of visualizing and quantifying the inter-company transaction network into a graph structure.

[0108] In this case, the edges between the central node and neighboring nodes signify the existence of transaction history, and the strength of the edges is weighted using values ​​such as transaction amount, frequency, and item similarity.

[0109] With respect to the first network graph above, the step of calculating edge weights corresponding to each edge based on transaction amount, transaction frequency, and transaction item similarity is a process of quantifying the qualitative characteristics of transactions beyond simple connection relationships.

[0110] For example, if a transaction of 100 million won is repeated 10 or more times a month, the edge weight is set high, and additional similarity points may be assigned as the transaction items are similar to each other (e.g., same HS code).

[0111] The step of designating a peripheral node directly connected to only one of the aforementioned peripheral nodes, either the central node or another peripheral node, as a terminal node, and the step of tracing a first path, which is the shortest path connecting from the central node to each terminal node, and summing the weights of the edges included in each first path to calculate a first network length for each terminal node, are processes for determining the overall structure / size of the transaction network.

[0112] With respect to the first network graph above, in the step of calculating the first network average length, first network maximum length, first network minimum length, and first network length standard deviation based on the first network length, the average length, maximum length, minimum length, standard deviation, etc. are derived based on the first network lengths calculated for each path, and the characteristics / homogeneity of the network can be measured according to the process described below.

[0113] The steps of designating a peripheral node connected to three or more other peripheral nodes or connected to a central node and two or more other peripheral nodes as a nested node, and designating a peripheral node among the other peripheral nodes connected to each nested node that is not included in the first path of the nested node as a radiating node, are intended to identify a major point that causes a bottleneck or a diffusion path within the network.

[0114] Subsequently, by tracing the shortest path (second path) from the radiating node to the terminal node and summing the edge weights to calculate the second network length, the characteristics of the diffusion structure can be numerically measured.

[0115] For the first network graph above, the step of calculating network characteristic indicators, which calculates the length distribution homogeneity index, bottleneck index, and distribution bias index based on the first network average length, first network maximum length, first network minimum length, first network length standard deviation, second network average length, second network maximum length, second network minimum length, second network length standard deviation, and the number of radiating nodes, is a key process for quantifying network structural characteristics.

[0116] For example, the coefficient of variation of path length within a network indicates homogeneity, the concentration of radiating nodes indicates bottlenecks, and the bias coefficient measures structural asymmetry.

[0117] The structural determination step, which determines the structural properties of the first network graph as one of compact homogeneous type, bottleneck-dominant type, heterogeneous risk type, and biased type according to pre-set standard conditions for network characteristic indicators, is a step of classifying the network type through the calculated indicators.

[0118] For example, if the length distribution is homogeneous and the bottleneck indicator is low, it is determined to be a compact homogeneous type, and conversely, if connections are excessively concentrated on a specific node, it is determined to be a bottleneck-dominated type.

[0119] Based on the previously designated preference structure for the aforementioned first company, the step of extracting the third-third company, which is the third-second company that matches the first company among the third-second companies, is a process for recommending an optimal trading partner that fits the company-specific tendencies or strategic requirements.

[0120] For example, if Company A prefers a compact homogeneous structure, Company 3-2, which is determined to have the same structure, is designated as a candidate.

[0121] The step of designating the first network graph of the aforementioned third-third company as the second network graph, and the step of designating the list of first companies included in the second network graph as new business partner candidates, are the final processes for connecting the network structure determination results to actual business partner companies. Through this, it is possible to present not only companies with large transaction volumes but also companies with secured network stability as new business partner candidates.

[0122] In addition, the step of calculating the first centrality index comprises: a step of log-normalizing the total sum of transaction amounts for each of the 3-1 companies; a step of standardizing the transaction frequency for each of the 3-1 companies by dividing it by the average transaction frequency of the corresponding 1 group; a step of converting the number of sales outlets and the number of purchase outlets for each of the 3-1 companies into ratios relative to the maximum value within the corresponding 1 group, respectively; a step of converting credit rating information for each of the 3-1 companies into a credit index pre-specified by grade; a step of calculating the second centrality index by linearly summing the total sum of transaction amounts, transaction frequency, number of sales outlets, number of purchase outlets, and credit index according to pre-specified weights for each of the 3-1 companies included in the 1 group; and a step of calculating the third centrality index by normalizing the second centrality index calculated for each of the 3-1 companies included in the 1 group according to the interval between the minimum and maximum values ​​of all second centrality indices calculated for the corresponding 1 group. and a step of designating the third centrality index as the first centrality index corresponding to the 3-1 enterprise; wherein the step of calculating the network characteristic indicator comprises: a step of calculating a first coefficient of variation by dividing the first network average length by the first network length standard deviation; a step of calculating a second coefficient of variation by dividing the second network average length by the second network length standard deviation; a step of calculating a length distribution homogeneity index by weighted averaging the first coefficient of variation and the second coefficient of variation; a step of calculating the number of radial connections per radial node, which is the number of edges connected to each of the radial nodes; a step of calculating a first radial concentration by dividing the number of radial connections per radial node by the average value of the number of radial connections per radial node; a step of calculating a second radial concentration, which is 'the number of radial nodes where the first radial concentration is greater than or equal to a predetermined threshold'; and a step of calculating a bottleneck indicator, which is the value obtained by dividing the second radial concentration by the number of surrounding nodes.The method may include: a step of calculating a first bias coefficient of '(maximum length of the first network + minimum length of the first network - 2 × average length of the first network) / (maximum length of the first network - minimum length of the first network)'; a step of calculating a second bias coefficient of '(maximum length of the second network + minimum length of the second network - 2 × average length of the second network) / (maximum length of the second network - minimum length of the second network)'; a step of calculating a first bias weight of 'number of terminal nodes / (number of terminal nodes + number of radiating nodes)'; a step of calculating a second bias weight of 'number of radiating nodes / (number of terminal nodes + number of radiating nodes)'; and a step of calculating a bias index of the distribution of 'first bias coefficient × first bias weight + second bias coefficient × second bias weight'.

[0123] The step of calculating the first centrality index mentioned above is a process of quantifying the transaction importance of individual firms by evaluating them in a multidimensional manner.

[0124] In this stage, an index is calculated by reflecting transaction amount, transaction frequency, diversity of suppliers and buyers, and credit rating information, and a standardized centrality index is derived through relative comparison within the group.

[0125] The step of log-normalizing the total transaction amount for each company in Section 3-1 above is a process that reduces bias caused by cases where a specific company's transaction amount is excessively large or small, and enables balanced reflection during index calculation.

[0126] For example, if Company A's annual total transaction volume is 10 billion won and Company B's annual total transaction volume is 100 million won, Company A's influence appears excessively large when simply aggregated. By applying a log transformation to correct this, the gap between companies with large and small transaction volumes can be mitigated, thereby increasing the stability of the index.

[0127] The step of standardizing the transaction frequency for each company in Section 3-1 above by dividing it by the average transaction frequency of the corresponding Group 1 is a process of adjusting so that the relative difference between companies with a high number of transactions and those with a low number of transactions can be compared using the same standard.

[0128] For example, if the average number of transactions within a group is 20 and Company A performs 40 transactions, the standardized transaction frequency value is calculated as 2.

[0129] The step of converting the number of sales and purchasing partners for each of the above-mentioned companies into ratios relative to the maximum value within the corresponding Group 1 is a process of evaluating how diverse the business partners a company is connected to.

[0130] For example, if the maximum number of sales outlets within a group is 50 and a specific company has 25 sales outlets, the sales outlet diversity index is calculated as 0.5. The number of suppliers is calculated in the same way.

[0131] The step described in Section 3-1 above, which converts credit rating information by company into a pre-designated credit index by grade, is a process of quantifying the credit ratings provided by external rating agencies and reflecting them in the index calculation.

[0132] For example, grade-specific weights can be pre-set, such as 1.0 for AAA, 0.9 for AA, and 0.8 for A.

[0133] For each of the 3-1 companies included in the above 1st group, the step of calculating the 2nd centrality index by linearly summing the total transaction amount, transaction frequency, number of sales outlets, number of purchase outlets, and credit index according to pre-specified weights is a process of integrating the previously calculated individual indicators to calculate a single index.

[0134] For example, if weights of 0.3 are assigned to the transaction amount, 0.25 to the transaction frequency, 0.15 to the number of sales outlets and purchase outlets respectively, and 0.15 to the credit index, and the results are summed, the second centrality index for each firm is calculated.

[0135] The step of calculating a third centrality index by normalizing the second centrality index calculated for each 3-1 firm included in the first group according to the interval between the minimum and maximum values ​​of all second centrality indices calculated for the first group is a process that converts the centrality index between firms into a value between 0 and 1 to enable relative comparison.

[0136] For example, if the minimum value of the second centrality index is 0.2 and the maximum value is 0.9, a firm with a second centrality index of 0.5 has a third centrality index of (0.5-0.2) / (0.9-0.2) = 0.4286.

[0137] The step of designating the aforementioned third centrality index as the first centrality index corresponding to the 3-1 company is a process of setting the finally calculated relative centrality index as the representative value of the company. Through this, the relative importance of all companies within the same group can be objectively compared, and this can subsequently be utilized to select new client candidates or for network analysis.

[0138] The above-mentioned step of calculating network characteristic indicators is a series of processes for quantifying, comparing, and evaluating the structural characteristics of the first network and the second network. In this step, the final indicator is calculated by comprehensively reflecting the homogeneity of the length distribution, the presence of bottleneck structures, and the bias of the distribution.

[0139] The step of calculating the first coefficient of variation by dividing the first network average length by the first network length standard deviation is defined as the value obtained by dividing the average value of the path lengths reaching the terminal node in the first network by the length standard deviation.

[0140] The coefficient of variation indicates the relative stability of the mean to the variance, and a higher value means that the path lengths are more evenly distributed.

[0141] The step of calculating the second coefficient of variation by dividing the average length of the second network by the standard deviation of the second network length is calculated in the same way for the second network. Since the second network is calculated based on the path from the radiating node to the terminal node, it serves to measure the homogeneity of the extension structure derived from the center.

[0142] The step of calculating the length distribution homogeneity index by weighted averaging the first coefficient of variation and the second coefficient of variation is a process of integrating the coefficients of variation of the two networks into a single indicator. Here, the weights can be pre-set according to the relative importance or the ratio of the number of nodes of the first network and the second network. For example, if the number of nodes in the first network is 70% of the total and the number of nodes in the second network is 30%, a weighted average can be performed according to that ratio.

[0143] The step of calculating the number of radial connections per radial node, which is the number of edges connected to each of the aforementioned radial nodes, is a process of quantifying how many edges each radial node derived from an overlapping node is connected to a neighboring node.

[0144] The step of calculating the first radiation concentration by dividing the number of radiation connections per radiation node by the average value of the number of radiation connections per radiation node indicates how intensive the connectivity of each radiation node is compared to the average. For example, if the average number of connections is 4 and a specific radiation node has 8 connections, the first radiation concentration of that radiation node is calculated as 2.

[0145] The step of calculating the second radiation concentration, which is the number of radiation nodes whose first radiation concentration is greater than or equal to a predetermined threshold, is a process of aggregating the number of nodes whose concentration exceeds a certain standard and which are likely to cause a bottleneck phenomenon.

[0146] The step of calculating the bottleneck index, which is the value obtained by dividing the second radiation concentration by the number of surrounding nodes, represents the proportion of nodes that can act as bottlenecks within the entire network. A higher value indicates an unbalanced structure where a small number of nodes monopolize an excessive number of connections.

[0147] The step of calculating a first bias coefficient by dividing (first network maximum length + first network minimum length - 2 times first network average length) by (first network maximum length - first network minimum length) using the first network maximum length, first network minimum length, and first network average length is a process of measuring whether the first network length distribution is symmetrical around the mean.

[0148] The step of calculating the second bias coefficient in the same manner using the second network maximum length, second network minimum length, and second network average length above indicates the bias of the second network length distribution.

[0149] The step of calculating a first bias weight by dividing the number of terminal nodes by the total number of nodes (number of terminal nodes + number of radiating nodes) using the above-mentioned number of terminal nodes and number of radiating nodes represents the proportion occupied by terminal nodes in the network.

[0150] The step of calculating the second bias weight by dividing the number of radiating nodes by the total number of nodes (number of terminal nodes + number of radiating nodes) reflects the proportion of radiating nodes that cause structural imbalance in the network.

[0151] The step of calculating the distribution bias index by multiplying the first bias coefficient by the first bias weight and the second bias coefficient by the second bias weight, and then summing them, is a process of quantifying the asymmetry and structural bias of the length distribution in an integrated manner. Consequently, a larger value indicates that the network is skewed in a specific direction, while a smaller value indicates a balanced structure.

[0152] An apparatus according to one embodiment includes a processor and memory. The processor may include at least one apparatus described above through the drawings or perform at least one method described above through the drawings. The memory may store information related to the method described above or store a program in which the method described above is implemented. The memory may be volatile memory or non-volatile memory.

[0153] The processor can execute a program and control the device. The code of the program executed by the processor can be stored in memory. The device can be connected to an external device (e.g., a personal computer or a network) through an input / output device (not shown in the drawing) and exchange data.

[0154] The embodiments described above may be implemented as hardware components, software components, and / or combinations of hardware and software components. For example, the devices, methods, and components described in the embodiments may be implemented using one or more general-purpose or special-purpose computers, such as, for example, a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable gate array (FPGA), a programmable logic unit (PLU), a microprocessor, or any other device capable of executing and responding to instructions. The processing unit may execute an operating system (OS) and one or more software applications executed on said operating system. Additionally, the processing unit may access, store, manipulate, process, and generate data in response to the execution of the software. For ease of understanding, the processing unit may be described as being used as a single unit, but those skilled in the art will understand that the processing unit may include multiple processing elements and / or multiple types of processing elements. For example, the processing unit may include multiple processors or one processor and one controller. Additionally, other processing configurations, such as parallel processors, are also possible.

[0155] The method according to the embodiment may be implemented in the form of program instructions that can be executed through various computer means and recorded on a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, etc., either alone or in combination. The program instructions recorded on the medium may be those specifically designed and configured for the embodiment, or they may be those known and available to those skilled in the art of computer software. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tapes; optical recording media such as CD-ROMs and DVDs; magneto-optical media such as floptical disks; and hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, and flash memory. Examples of program instructions include machine code, such as that generated by a compiler, as well as high-level language code that can be executed by a computer using an interpreter, etc. The hardware devices described above may be configured to operate as one or more software modules to perform the operation of the embodiment, and vice versa.

[0156] Software may include computer programs, code, instructions, or a combination of one or more of these, and may configure a processing unit to operate as desired or command the processing unit independently or collectively. Software and / or data may be permanently or temporarily embodied in any type of machine, component, physical device, virtual equipment, computer storage medium, or device so as to be interpreted by the processing unit or to provide instructions or data to the processing unit. Software may be distributed over networked computer systems and may be stored or executed in a distributed manner. Software and data may be stored on one or more computer-readable recording media.

[0157] Although the embodiments have been described above with reference to the limited drawings, those skilled in the art can apply various technical modifications and variations based on the above. For example, suitable results may be achieved even if the described techniques are performed in a different order than described, and / or if the components of the described system, structure, device, circuit, etc. are combined or assembled in a form different from described, or replaced or substituted by other components or equivalents.

[0158] 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

[0159] 1 : Device (Server) 2 : User terminal 3 : Database S100: Steps for collecting corporate data S110: Step of receiving a tax invoice, transaction statement, or contract S120: Step of extracting transaction industry, transaction item, and transaction amount and designating them as first user data S130: A step of receiving transaction history information and designating it as second user data S140: Step for calculating the reliability score S150: A step of sending an information review request message to the user terminal. S160: Step of adding first user data and second user data to enterprise data S170: A step of receiving credit rating information and corporate registration information corresponding to each of the first companies and adding them to corporate data. S200: Step of classifying first-order firms into multiple first-order groups S210: Preprocessing step S220: Step to generate the first vector S230: Step of clustering the first firms into the first group S300: The stage of selecting new client candidates S400: A step of transmitting new business partner candidate information to the user terminal of the first company. S500: A step of transmitting information of the first company corresponding thereto to the user terminal of a new business partner candidate. S600: Step of receiving search conditions S700: Step to extract the second company S800: Step for calculating movement paths between delivery locations of the second companies S900: A step of transmitting information of the second company and movement paths between delivery locations.

Claims

Claim 1 A method for providing a solution for corporate grouping and network construction within a group through artificial intelligence-based corporate data analysis, wherein the device includes a processor, memory, a communication module, and a non-transient storage medium, and the processor executes a program stored in the non-transient storage medium, the method comprising: collecting corporate data, which is data of a first company, from a user terminal and a corporate information API; inputting the corporate data into an artificial intelligence model to classify the first companies into a plurality of first groups; mutually analyzing suppliers and customers for each first group to select new business partner candidates, including new suppliers or new customers, for each first company; transmitting information on the new business partner candidates to the user terminal of the first company for which the new business partner candidates have been selected; transmitting information on the first company corresponding thereto to the user terminal of the new business partner candidates; receiving search conditions from the user terminal; extracting a second company, which is a first company corresponding to the search conditions, among the first companies; and calculating the movement path between delivery locations of the second companies based on the information of the second company. and the step of transmitting information of the second company and the route between the delivery location to the user terminal; wherein the step of collecting the company data comprises: receiving a tax invoice, a transaction statement, or a contract from the user terminal; extracting the transaction industry, transaction items, and transaction amount from the tax invoice, transaction statement, or contract based on an OCR module and designating them as first user data; receiving transaction history information including sales source information, purchase source information, transaction industry, transaction items, and transaction amount from the user terminal and designating it as second user data; comparing the first user data and the second user data to calculate a reliability score based on the content match rate; and if the reliability score is below a predetermined threshold, transmitting an information re-examination request message to the user terminal.The method comprises: a step of adding the first user data and the second user data to the corporate data when the above reliability score is above a predetermined threshold; and a step of receiving credit rating information and corporate registration information corresponding to each of the first companies from the corporate information API and adding them to the corporate data; and the step of classifying the first companies into a plurality of first groups based on the corporate data includes: a preprocessing step of converting industry names, product names, and transaction item names included in the corporate data into standardized keywords through a natural language processing process including morphological analysis, stop word removal, and synonym dictionary mapping; a step of generating a first vector by vectorizing the standardized keywords extracted for each of the first companies; and a step of clustering the first companies into first groups based on the first vectors; and the step of selecting new business partner candidates includes: a step of calculating a first centrality index for each of the 3-1 companies included in the first group based on transaction amount, transaction frequency, number of sales outlets, number of purchase outlets, and credit rating information for each of the 1 groups; and a step of designating the 3-1 company whose first centrality index is above a predetermined threshold as the 3-2 company. A step of generating a first network graph for each of the above 3-2 companies by designating each of the above 3-2 companies as a central node and designating a 1 company that has a transaction history with the 3-2 company as a peripheral node; a step of calculating an edge weight corresponding to each edge for the above 1 network graph based on transaction amount, transaction frequency, and transaction item similarity; a step of designating a peripheral node directly connected to only one of the above peripheral nodes, either the central node or another peripheral node, as a terminal node; a step of tracing a first path, which is the shortest path connecting from the central node to each terminal node, and calculating a first network length for each terminal node by summing the edge weights of the edges included in each first path.For the first network graph above, a step of calculating the first network average length, the first network maximum length, the first network minimum length, and the first network length standard deviation based on the first network length; a step of designating a neighbor node connected to three or more other neighbor nodes or connected to a 'central node and two or more other neighbor nodes' as an overlapping node; for each overlapping node, a step of designating a 'neighbor node among the other connected neighbor nodes that is not included on the first path of the overlapping node' as a radiating node; a step of tracing a second path, which is the shortest path connecting from the radiating node to the terminal node, and calculating the second network length for each radiating node by summing the edge weights of the edges included in each second path; and for the first network graph above, a step of calculating the second network average length, the second network maximum length, the second network minimum length, and the second network length standard deviation based on the second network length. A step for calculating network characteristic indicators, wherein for the first network graph, a length distribution homogeneity index, a bottleneck indicator, and a distribution bias index are calculated based on the first network average length, the first network maximum length, the first network minimum length, the first network length standard deviation, the second network average length, the second network maximum length, the second network minimum length, the second network length standard deviation, and the number of radiating nodes; a structural determination step, wherein the structural properties of the first network graph are determined as one of compact homogeneous type, bottleneck dominant type, heterogeneous risk type, and bias type according to pre-set standard conditions for the network characteristic indicators; a step for extracting a third-third company, which is a third-second company that matches the first company among third-second companies, based on a pre-specified preference structure for the first company; a step for designating the first network graph of the third-third company as the second network graph; and a step for designating the list of first companies included in the second network graph as new business partner candidates.A method for providing a solution for corporate grouping and building intra-group networks through AI-based corporate data analysis, including Claim 2 delete Claim 3 delete Claim 4 A method for providing a solution for corporate grouping and network construction within a group through AI-based corporate data analysis, further comprising: a step of, after the step of calculating the movement path between the delivery locations of the second companies, automatically listing component suppliers and sub-process partner companies required for the production of the finished product based on the item and process flow of the finished product that the user intends to produce. Claim 5 A method for providing a solution for corporate grouping and network establishment within a group through AI-based corporate data analysis, further comprising the step of optimizing the entire supply chain route by combining location information, transportation distance, estimated lead time, and cost data of the listed partner companies in claim 4. Claim 6 In claim 5, the step of optimizing the entire supply chain route comprises: simulating the supply chain route based on the final delivery location and automatically calculating a location placement plan suitable for the supply chain flow according to the simulation results; a method for providing a solution for corporate grouping and network construction within a group through AI-based corporate data analysis. Claim 7 In claim 6, the step of optimizing the entire supply chain route comprises: a method for providing a solution for corporate grouping and building a network within a group through AI-based corporate data analysis, wherein, when logistics cooperation between tenant companies within the same building or complex is required, truck entry control, freight elevator calling, and warehouse loading and dispatch scheduling are automatically controlled. Claim 8 A method for providing a solution for corporate grouping and building a network within a group through AI-based corporate data analysis, further comprising the step of automatically performing escrow payment or point / coin-based step-by-step payment and settlement according to contract conditions and the progress of process stages in claim 7.

Images