An intelligent home personalization design system and method thereof

Abstract

The application discloses an intelligent home personalization design system and method, and relates to the technical field of intelligent home design.The method extracts features from a house type data set, constructs structural similarity between any two house types, and quantitatively analyzes environmental factors of the same type of house structure, calculates the space lighting coefficient and ventilation flow index of the i-th house type, and constructs the natural adaptation value of the i-th house type.Through accurate quantification of the lighting and ventilation performance of the house type, the differences in environmental adaptability of different house types can be scientifically reflected, and the shortcomings of traditional methods that only rely on structural area or simple layout matching and ignore environmental quality are solved.Combining with family members and personalized preferences of users, multi-dimensional personalized expression of design requirements is realized, and it is ensured that the recommended scheme not only meets the space function, but also better meets the actual living habits and comfort requirements of users.

Description

Technical Field

[0001] This invention relates to the field of smart home design technology, specifically to a smart home personalized design system and method. Background Technology

[0002] As people's living standards continue to improve, the demand for personalized living environments is growing, especially in the field of smart home design. Users are placing higher demands on environmental factors such as ventilation and natural lighting in their living spaces. In practical applications, traditional home design recommendations often rely on simple matching of room size and layout, neglecting the environmental adaptability of the apartment in terms of natural lighting and air circulation, and failing to fully consider the diverse composition and lifestyle preferences of family members. This lack of refined, multi-dimensional feature modeling leads to a significant discrepancy between recommended home design schemes and users' actual preferences, failing to effectively meet the comprehensive needs of modern families for comfort, health, and personalization.

[0003] Especially in urban residential buildings, due to the complex influencing factors such as building orientation, building spacing, and window layout, many apartment layouts, while structurally similar, exhibit significant differences in lighting conditions and ventilation performance. Traditional methods struggle to accurately measure and systematically analyze these differences, thus failing to provide scientifically sound design recommendations. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a smart home personalized design system and method to solve the problems mentioned in the background section.

[0005] To achieve the above objectives, the present invention provides a personalized smart home design method, comprising the following steps:

[0006] Step 1: Obtain the apartment layout data of the target family and store it in the apartment layout data group. Extract features from the apartment layout data group and construct any two apartment layouts. and Structural similarity between And form the apartment space feature matrix S;

[0007] Step 2: Based on the apartment type spatial feature matrix S, construct apartment type clusters using hierarchical clustering, perform structural similarity clustering analysis on all apartment types, and classify different types of residential space layouts.

[0008] Step 3: Quantitatively analyze the environmental factors of the same type of apartment structure and calculate the spatial daylighting coefficient of the i-th apartment type. and ventilation index Construct the natural fit value for the i-th apartment type In addition, by combining the member composition and preference tags of the target family, a user preference vector Pp is constructed to form a design requirement data set;

[0009] Step 4: Based on the design requirements data set, use the weighted similarity matching method to calculate the design preference matching ratio M between the natural fit value of the i-th apartment type and the user preference vector Pp, and output several successfully matched apartment types. Sort all apartment types and preset the matching degree threshold X. Select several apartment types with a design preference matching ratio M higher than X as the final recommended scheme.

[0010] Preferably, the floor plan CAD drawing and 3D scanning results of the user's residence are obtained through an intelligent surveying terminal;

[0011] Intelligent mapping terminals include lidar equipment or panoramic scanners;

[0012] Structured data of different residential units are extracted, including the number of rooms, layout area, opening orientation, window area, and floor orientation, to construct the feature set of the i-th unit type. The set of apartment type features of the j-th apartment type ;

[0013] ;

[0014] ;

[0015] in, This represents the room distribution vector of the i-th apartment type. Let i represent the area vector of the i-th apartment type. Let represent the opening direction vector of the i-th apartment type. This represents the ratio of the window area to the total area of ​​the i-th apartment unit. This represents the floor height and orientation of the i-th apartment unit.

[0016] This represents the room distribution vector of the j-th apartment type. Let the area vector of the j-th apartment type be represented. Let represent the opening direction vector of the j-th apartment type. This represents the ratio of the window area to the unit type j. Represent the floor height and orientation attributes of the j-th apartment type; represent the set F of all apartment type features. i The set of apartment type features of the j-th apartment Store the apartment type data set H={F1, F2, ..., F n}; n represents the total number of apartment types;

[0017] Calculate the similarity between any two apartment layouts using graphical analysis and spatial similarity formulas. and Structural similarity between ;

[0018]

[0019] in, Let represent the Euclidean distance between the sets of feature values ​​of the i-th and j-th apartment types. express and cosine similarity, and Represented as weights, , ; ;

[0020] Based on structural similarity Construct the apartment layout spatial feature matrix S;

[0021]

[0022] In this matrix, the diagonal is 0, the similarity between matrices is constant, and the upper and lower triangles are symmetrical. Normalization and redundancy removal are then performed to obtain the processed apartment layout spatial feature matrix. .

[0023] Preferably, the processed apartment space feature matrix Input a clustering analysis algorithm, construct apartment type clusters using hierarchical clustering, perform structural similarity clustering analysis on all apartment types, classify the apartment types within each cluster into classes with the same spatial structure, forming different residential space types, and construct an apartment type cluster set C; C = {C1, C2, ..., C6}. k}; where each cluster corresponds to a spatial structure type; and the structure class label of the k-th type of house is marked.

[0024] Preferably, the spatial daylighting coefficient of the i-th unit type and ventilation index The construction method is as follows:

[0025] For the same type of apartment, extract window-related information for each room, including window location coordinates, orientation angle, and distribution of external obstructions.

[0026] Based on geographic location information and shading models, and considering the orientation of room windows, the Radiance lighting simulation tool is used to calculate the sunshine duration Ts and maximum daylight intensity Ls for the corresponding apartment type, thus forming the spatial daylight coefficient for the i-th apartment type. ;

[0027] ;

[0028] In the formula, This indicates the weight of the sunshine duration Ts for the corresponding type of apartment. This indicates the weight of the maximum daylight intensity Ls for the corresponding apartment type. and All are constants, and ;

[0029] It also extracts the door and window distribution information of each room, including the size, location, opening area, and connection with the outside or adjacent rooms. Based on the spatial layout, it analyzes the unobstructedness of the ventilation path and identifies the natural convection path, the angle between the air intake and exhaust, and the area A of each opening.

[0030] Natural convection pathways include,

[0031] In a north-south facing apartment, the south-facing windows bring in air while the north-facing windows exhaust air, creating the first cross ventilation path.

[0032] The vertical convection formed by the height difference of the openings includes the thermal pressure driven convection path formed by the upper window and the lower air inlet, which causes hot air to rise and cold air to enter, forming a second vertical ventilation path.

[0033] The connection between semi-open spaces such as balconies, terraces, and courtyards and interior doors and windows forms a third ventilation path;

[0034] The inner courtyard or atrium assists in the vertical circulation of airflow, creating a "chimney effect" that drives the air to flow up and down, forming a fourth ventilation path;

[0035] Obtain the natural convection path, door and window convection angles, and opening areas to calculate the ventilation index of the i-th unit type. ;

[0036]

[0037] Where m represents the total number of identified natural convection paths. This represents the angle between the air intake and exhaust points of the j-th path. Let the effective opening area of ​​the j-th path be denoted as . This represents the path smoothness coefficient of the j-th path;

[0038]

[0039] in, This represents the total number of turns and bends along the j-th path. D represents the correction factor, where D=0.5 means that each turn reduces traffic flow by 0.5 units.

[0040] The preferred natural fit value for the i-th apartment type The calculation formula is as follows:

[0041] Combining the spatial lighting coefficient of the i-th unit and ventilation index Calculate the natural fit value of the i-th apartment type. :

[0042]

[0043] In the formula, and Let represent the daylighting coefficients of the i-th apartment type, respectively. and ventilation index The weight, ; and These represent the maximum daylight intensity and maximum ventilation capacity observed in the same type of apartment, respectively, and are used for normalization.

[0044] Preferably, the steps for constructing the user preference vector Pp are as follows:

[0045] Collect users' preferences for sunlight in their living environment. If users prefer ample sunlight, increase their "sunlight preference" score by 3 points; if they prefer soft lighting, adjust their "sunlight preference" score to a moderate level and increase it by 1-2 points.

[0046] Collect users' preferences for sunlight. If a user prefers plenty of sunlight, increase their "sunshine preference" score by 3 points.

[0047] If the user prefers moderate lighting, then the "sunshine preference" rating will be set to a medium score of 2 points.

[0048] If the user prefers a soft or low-light environment, the "sunshine preference" score will be set to a low value of 1 point.

[0049] Collect users' preferences for ventilation. If users prefer good ventilation, set the "ventilation preference" score to a high value of 3 points.

[0050] If the user prefers moderate ventilation, the "ventilation preference" score will be set to a medium score of 2 points.

[0051] If a user prefers a quiet environment with little ventilation, the "ventilation preference" score will be set to a low value of 1 point.

[0052] By combining multiple user preferences, a user preference vector Pp is formed.

[0053] The preferred approach to constructing a comprehensive smart home design model is as follows:

[0054] Collect a historical matching case library and create input-output sample pairs for each case;

[0055] First, standardize the ratings in the user preference vector Pp to ensure the rating data is within a uniform dimensional range; second, standardize the natural fit value of the i-th apartment type. Dimensionless normalization is performed to ensure that the lighting and ventilation indicators match the dimensions preferred by users.

[0056] Then, using a weighted similarity matching method, the design preference matching ratio M between the natural fit value of the i-th apartment type and the user preference vector Pp is calculated to output several successfully matched apartment types.

[0057] The system employs decision tree or deep learning model algorithms to construct and train a large-scale smart home design model. During training, cross-validation and early stopping mechanisms are introduced to prevent overfitting. After training, the model is applied to the test set to calculate the design preference matching ratio M, thereby outputting several successfully matched house types.

[0058] All apartment types are sorted, and a matching threshold X is preset. Several apartment types with a design preference matching ratio M higher than X are selected as the final recommended solutions.

[0059] Preferably, the design preference matching ratio M is obtained in the following way:

[0060]

[0061] in, This represents the normalized natural fit value for the i-th apartment type. Let f represent the f-th rating component in the user preference vector, where e is the vector dimension;

[0062] The closer the design preference matching ratio M is to 1, the higher the degree of matching between the i-th apartment type and the user's preferences.

[0063] Preferably, the final recommended solution is obtained in the following way:

[0064] Based on the design preference matching ratio M, the design output with the highest matching degree is selected; the output design includes whole house style suggestions, space layout diagram, equipment suggestion list and construction budget estimate.

[0065] A smart home personalized design system, comprising,

[0066] The apartment layout data acquisition module is used to obtain residential structure diagrams, orientation and window opening information, and to construct apartment layout data groups;

[0067] The spatial analysis module is used to extract lighting and ventilation information to form the natural fit value for the i-th unit type. ;

[0068] The preference modeling module collects user member information, behavioral preferences, and budget expectations to generate a user preference vector Pp.

[0069] The design includes a large model building module for collecting historical matching case libraries and establishing input-output sample pairs for each case. The module standardizes the ratings in the user preference vector Pp to ensure the rating data is within a uniform dimensional range. Secondly, it calculates the natural fit value of the i-th apartment type. Dimensionless normalization is performed to ensure that the lighting and ventilation indicators match the dimensions preferred by users.

[0070] The model optimization and recommendation module is used to calculate the design preference matching ratio M, and preset the matching degree threshold X, and select several house types with a design preference matching ratio M higher than X as the final recommended solutions.

[0071] This invention provides a smart home personalized design system and method. It has the following beneficial effects:

[0072] By introducing a two-way modeling mechanism that combines natural fit values ​​and user preference vectors, this system breaks through the limitations of traditional apartment recommendation systems that rely solely on structural features or area dimensions. It can quantify the actual performance differences of apartment types in terms of environmental factors such as natural lighting and ventilation from multiple dimensions, making it particularly suitable for scenarios in densely populated urban residential areas where the structures are similar but the environmental adaptability varies significantly.

[0073] Furthermore, by modeling the composition of the target family members and their lifestyle preference tags, a personalized user preference vector Pp is constructed, enabling the design scheme to shift from "spatial adaptation" to "human-living integration". Then, by using a weighted similarity matching algorithm (including cosine similarity, inverse Euclidean distance, or weighted Manhattan distance) to accurately match the natural adaptation value with the user preference vector, the user satisfaction and living comfort of the recommended scheme can be significantly improved. Attached Figure Description

[0074] Figure 1 This is a schematic diagram illustrating the steps of a smart home personalized design method according to the present invention;

[0075] Figure 2 This is a schematic diagram of the process of a smart home personalized design system according to the present invention. Detailed Implementation

[0076] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0077] Example 1

[0078] Please see Figure 1 This invention provides a method for personalized smart home design, comprising the following steps:

[0079] Step 1: Obtain the apartment layout data of the target family and store it in the apartment layout data group. Extract features from the apartment layout data group and construct any two apartment layouts. and Structural similarity between And form the apartment space feature matrix S;

[0080] Step 2: Based on the apartment type spatial feature matrix S, construct apartment type clusters using hierarchical clustering, perform structural similarity clustering analysis on all apartment types, and classify different types of residential space layouts.

[0081] Step 3: Quantitatively analyze the environmental factors of the same type of apartment structure and calculate the spatial daylighting coefficient of the i-th apartment type. and ventilation index Construct the natural fit value for the i-th apartment type In addition, by combining the member composition and preference tags of the target family, a user preference vector Pp is constructed to form a design requirement data set;

[0082] Step 4: Based on the design requirements data set, use the weighted similarity matching method to calculate the design preference matching ratio M between the natural fit value of the i-th apartment type and the user preference vector Pp, and output several successfully matched apartment types. Sort all apartment types and preset the matching degree threshold X. Select several apartment types with a design preference matching ratio M higher than X as the final recommended scheme.

[0083] In this embodiment, the present invention breaks through the limitations of traditional housing recommendation systems that rely solely on structural features or area dimensions by introducing a two-way modeling mechanism of natural adaptation value and user preference vector. It can quantitatively measure the actual performance differences of housing types in environmental factors such as natural lighting and ventilation from multiple dimensions, and is particularly suitable for scenarios in densely populated urban residential areas where the structures are similar but the environmental adaptability is significantly different.

[0084] Furthermore, by modeling the composition of the target family members and their lifestyle preference tags, a personalized user preference vector Pp is constructed, enabling the design scheme to shift from "spatial adaptation" to "human-living integration". Then, by using a weighted similarity matching algorithm (including cosine similarity, inverse Euclidean distance, or weighted Manhattan distance) to accurately match the natural adaptation value with the user preference vector, the user satisfaction and living comfort of the recommended scheme can be significantly improved.

[0085] Example 2

[0086] This embodiment is an explanation of Embodiment 1. Specifically, the CAD drawing of the user's residential floor plan and the 3D scanning results are obtained through an intelligent surveying terminal.

[0087] Intelligent mapping terminals include lidar equipment or panoramic scanners;

[0088] Structured data of different residential units are extracted, including the number of rooms, layout area, opening orientation, window area, and floor orientation, to construct the feature set of the i-th unit type. The set of apartment type features of the j-th apartment type ;

[0089] ;

[0090] ;

[0091] in, This represents the room distribution vector of the i-th apartment type. Let i represent the area vector of the i-th apartment type. Let represent the opening direction vector of the i-th apartment type. This represents the ratio of the window area to the total area of ​​the i-th apartment unit. This represents the floor height and orientation of the i-th apartment unit.

[0092] This represents the room distribution vector of the j-th apartment type. Let the area vector of the j-th apartment type be represented. Let represent the opening direction vector of the j-th apartment type. This represents the ratio of the window area to the unit type j. Represent the floor height and orientation attributes of the j-th apartment type; represent the set F of all apartment type features. i The set of apartment type features of the j-th apartment Store the apartment type data set H={F1, F2, ..., F n}; n represents the total number of apartment types;

[0093] Calculate the similarity between any two apartment layouts using graphical analysis and spatial similarity formulas. and Structural similarity between ;

[0094]

[0095] in, Let represent the Euclidean distance between the sets of feature values ​​of the i-th and j-th apartment types. express and cosine similarity, and Represented as weights, , ; ;

[0096] Based on structural similarity Construct the apartment layout spatial feature matrix S;

[0097]

[0098] In this matrix, the diagonal is 0, the similarity between matrices is constant, and the upper and lower triangles are symmetrical. Normalization and redundancy removal are then performed to obtain the processed apartment layout spatial feature matrix. .

[0099] In this embodiment, the present invention acquires detailed floor plan data of user residences through intelligent mapping terminals (such as lidar devices or panoramic scanners), and combines CAD drawings and three-dimensional spatial models to achieve structured extraction of floor plan features, covering key parameters such as room distribution, area layout, window ratio and orientation; furthermore, it quantifies the structural similarity between different floor plans through a spatial similarity algorithm (a weighted calculation that integrates Euclidean distance and cosine similarity) to construct a standardized floor plan spatial feature matrix, which can effectively improve the accuracy of floor plan classification and design recommendations.

[0100] Example 3

[0101] This embodiment is an explanation of Embodiment 1. Specifically, it describes the processed apartment space feature matrix. Input a clustering analysis algorithm, construct apartment type clusters using hierarchical clustering, perform structural similarity clustering analysis on all apartment types, classify the apartment types within each cluster into classes with the same spatial structure, forming different residential space types, and construct an apartment type cluster set C; C = {C1, C2, ..., C6}. k}; where each cluster corresponds to a spatial structure type; and the structure class label of the k-th type of house is marked.

[0102] In this embodiment, the normalized apartment layout spatial feature matrix is ​​input into a hierarchical clustering analysis algorithm to achieve structural similarity clustering of all apartment layouts. This automatically identifies apartment layouts with similar spatial arrangements and groups them into the same cluster, constructing a set of apartment layout clusters C with various spatial structure types. This method effectively uncovers potential commonalities in spatial structure, improving the systematicness and objectivity of apartment layout classification. Furthermore, by labeling each cluster with structural tags, it enables standardized definitions and rapid indexing of different types of residential spaces.

[0103] Example 4

[0104] This embodiment is an explanation based on Embodiment 1. Specifically, the spatial daylighting coefficient of the i-th unit type... and ventilation index The construction method is as follows:

[0105] For the same type of apartment, extract window-related information for each room, including window location coordinates, orientation angle, and distribution of external obstructions.

[0106] Based on geographic location information and shading models, and considering the orientation of room windows, the Radiance lighting simulation tool is used to calculate the sunshine duration Ts and maximum daylight intensity Ls for the corresponding apartment type, thus forming the spatial daylight coefficient for the i-th apartment type. ;

[0107] ;

[0108] In the formula, This indicates the weight of the sunshine duration Ts for the corresponding type of apartment. This indicates the weight of the maximum daylight intensity Ls for the corresponding apartment type. and All are constants, and ;

[0109] It also extracts the door and window distribution information of each room, including the size, location, opening area, and connection with the outside or adjacent rooms. Based on the spatial layout, it analyzes the unobstructedness of the ventilation path and identifies the natural convection path, the angle between the air intake and exhaust, and the area A of each opening.

[0110] Natural convection pathways include,

[0111] In a north-south facing apartment, the south-facing windows bring in air while the north-facing windows exhaust air, creating the first cross ventilation path.

[0112] The vertical convection formed by the height difference of the openings includes the thermal pressure driven convection path formed by the upper window and the lower air inlet, which causes hot air to rise and cold air to enter, forming a second vertical ventilation path.

[0113] The connection between semi-open spaces such as balconies, terraces, and courtyards and interior doors and windows forms a third ventilation path;

[0114] The inner courtyard or atrium assists in the vertical circulation of airflow, creating a "chimney effect" that drives the air to flow up and down, forming a fourth ventilation path;

[0115] By accurately identifying the cross ventilation path, vertical thermal pressure drive path, semi-open connection path, and the "chimney effect" path of the inner courtyard in the north-south facing apartment, we can comprehensively evaluate the air circulation efficiency of the apartment under different natural conditions, accurately depict the spatial distribution and strength changes of ventilation capacity, which not only helps to optimize window opening strategies and spatial layout, improve the utilization rate of natural ventilation, and reduce energy consumption, but also effectively improve indoor air quality and thermal comfort.

[0116] Obtain the natural convection path, door and window convection angles, and opening areas to calculate the ventilation index of the i-th unit type. ;

[0117]

[0118] Where m represents the total number of identified natural convection paths. This represents the angle between the air intake and exhaust points of the j-th path. Let the effective opening area of ​​the j-th path be denoted as . This represents the path smoothness coefficient of the j-th path;

[0119]

[0120] in, This represents the total number of turns and bends along the j-th path, where D represents the correction factor. D=0.5 means that each turn reduces the flow by 0.5 units; for example, a fully continuous, straight ventilation path. =1; There is a corner =0.67; There were two turns. =0.5.

[0121] In this embodiment, the technology systematically extracts window and door structure information from sample apartment types of the same category, and combines this with the Radiance lighting simulation tool to accurately calculate sunshine duration and maximum light intensity, constructing a spatial daylighting coefficient. This allows for a more scientific quantification of the natural lighting performance of the apartment type. Simultaneously, based on the distribution of doors and windows and the spatial structure, it identifies multiple natural convection ventilation paths and constructs a ventilation circulation index covering key elements such as convection angle, opening area, and path unobstructedness. This enables a multi-dimensional and refined evaluation of the natural ventilation capacity of the apartment type, improving the comprehensive matching ability of the design scheme in terms of living comfort, energy efficiency, and ecological adaptability. This helps to more accurately select high-quality living space types that meet the preferences of target users.

[0122] Example 5

[0123] This embodiment is an explanation of embodiment 3. Specifically, the natural fit value of the i-th apartment type. The calculation formula is as follows:

[0124] Combining the spatial lighting coefficient of the i-th unit and ventilation index Calculate the natural fit value of the i-th apartment type. ,

[0125]

[0126] In the formula, and Let represent the daylighting coefficients of the i-th apartment type, respectively. and ventilation index The weight, ; and These represent the maximum observed natural light intensity and maximum ventilation capacity within the same type of apartment, used for normalization. The specific weights need to be determined based on actual application scenarios and user needs, through data analysis or expert experience. Under normal circumstances, and Each accounts for 50%, unless the weighting can be adjusted for specific climates or user preferences, for example.

[0127] In areas with ample sunlight, increase the weight of daylighting to 0.7 and decrease the weight of ventilation to 0.3;

[0128] In hot and humid regions, increase the ventilation weight to 0.7 and decrease the lighting weight to 0.3;

[0129] In this embodiment, the spatial daylighting coefficient and ventilation circulation index are introduced to comprehensively calculate the natural fit value of the i-th apartment type. By adopting normalization processing and weight adjustment mechanisms, a quantitative assessment of the natural daylighting and ventilation performance of different apartment types can be achieved, making the assessment results highly comparable and objective. At the same time, by combining the reference of the normalized maximum value, the differences in indicator scale caused by different apartment types such as area and orientation are effectively eliminated, improving the accuracy of the natural performance indicators and the generalization ability of the model. This provides a more scientific and adaptable natural environment foundation for subsequent apartment type selection and intelligent recommendation.

[0130] Example 6

[0131] This embodiment is an explanation based on Embodiment 1. Specifically, the steps for constructing the user preference vector Pp are as follows:

[0132] Collect users' preferences for sunlight in their living environment. If users prefer ample sunlight, increase their "sunlight preference" score by 3 points; if they prefer soft lighting, adjust their "sunlight preference" score to a moderate level and increase it by 1-2 points.

[0133] Collect users' preferences for sunlight. If a user prefers plenty of sunlight, increase their "sunshine preference" score by 3 points.

[0134] If the user prefers moderate lighting, then the "sunshine preference" rating will be set to a medium score of 2 points.

[0135] If the user prefers a soft or low-light environment, the "sunshine preference" score will be set to a low value of 1 point.

[0136] Collect users' preferences for ventilation. If users prefer good ventilation, set the "ventilation preference" score to a high value of 3 points.

[0137] If the user prefers moderate ventilation, the "ventilation preference" score will be set to a medium score of 2 points.

[0138] If a user prefers a quiet environment with little ventilation, the "ventilation preference" score will be set to a low value of 1 point.

[0139] If there are elderly people in the user's household, it indicates a need to enhance the accessibility and safety of the living environment, so the "Elderly Adaptability" score in the user preference vector will be increased by 2 points; if there are children in the user's household, it indicates a need to focus on children's safety and activity space, so the "Child-Friendliness" score will be increased by 2 points; if the user owns a pet, it indicates a need to consider pet activity and cleaning convenience, so the "Pet Adaptability" score will be increased by 1 point; if the user prefers a quiet environment, it indicates a high demand for noise control, so the "Quietness Requirement" score will be increased by 3 points; if the user has strong storage needs, it indicates a high requirement for storage space, so the "Storage Strength" score will be increased by 2 points; if the user has high smart home needs, it indicates a desire for a high degree of automation and intelligence in the home system, so the "..." score will be increased by 2 points. The "Intelligent Preference" score increases by 3 points; if a user prefers open space, indicating a preference for spacious and flowing layouts, the "Open Space" score increases by 1 point; if a user values ​​privacy, indicating a high demand for room privacy and partitions, the "Privacy Preference" score increases by 2 points; if a user values ​​convenient circulation, indicating a desire for a reasonable and convenient living flow, the "Circulation Convenience" score increases by 1 point; if a user is concerned about residential safety, indicating a high demand for safety protection measures, the "Safety Concern" score increases by 3 points; if a user values ​​health and comfort, indicating a high demand for air quality, lighting, and ventilation, the "Health and Comfort Sensitivity" score increases by 3 points. These factors combined form the user preference vector Pp.

[0140] In this embodiment, the user preference vector Pp construction steps can comprehensively capture users' multi-dimensional personalized needs for their living environment, covering multiple key factors such as sunlight and lighting preferences, ventilation preferences, and special needs of family members, making the design scheme closer to the user's real life and preferences. By refining the scoring system, it accurately reflects the user's emphasis on safety, comfort, intelligence, storage, and privacy, improving the targeting and practicality of the design recommendations. In addition, by combining family member structure and living habits, it enhances the adaptability of the apartment layout to different user groups, effectively avoiding the inaccurate recommendation problem caused by traditional single-dimensional matching.

[0141] Example 7

[0142] This embodiment is an explanation based on Embodiment 1. Specifically, the construction method of the large-scale smart home design model is as follows:

[0143] Collect a historical matching case library and create input-output sample pairs for each case;

[0144] First, standardize the ratings in the user preference vector Pp to ensure the rating data is within a uniform dimensional range; second, standardize the natural fit value of the i-th apartment type. Dimensionless normalization is performed to ensure that the lighting and ventilation indicators match the dimensions preferred by users.

[0145] Then, using a weighted similarity matching method, the design preference matching ratio M between the natural fit value of the i-th apartment type and the user preference vector Pp is calculated to output several successfully matched apartment types.

[0146] The system employs decision tree or deep learning model algorithms to construct and train a large-scale smart home design model. During training, cross-validation and early stopping mechanisms are introduced to prevent overfitting. After training, the model is applied to the test set to calculate the design preference matching ratio M, thereby outputting several successfully matched house types.

[0147] All apartment types are sorted, and a matching threshold X is preset. Several apartment types with a design preference matching ratio M higher than X are selected as the final recommended solutions.

[0148] The matching threshold X is derived from collecting satisfaction ratings from past users of the system for the recommended design schemes, and mapping these ratings to the corresponding matching ratios M. The lowest matching ratio corresponding to a satisfaction score ≥ 80 (or 4 stars or higher) is used as a reference. The matching threshold X typically ranges from 0.7 to 0.85 (70%–85%). When the matching ratio M between the i-th apartment type and the user's preference vector is ≥ 0.75, it is considered to meet basic personalization requirements and can be included in the recommendation set.

[0149] The design preference matching ratio M is obtained as follows:

[0150]

[0151] in, This represents the normalized natural fit value for the i-th apartment type. Let f represent the f-th rating component in the user preference vector, where e is the vector dimension;

[0152] The closer the design preference matching ratio M is to 1, the higher the degree of matching between the i-th apartment type and the user's preferences.

[0153] The final recommended solution is obtained as follows:

[0154] Based on the design preference matching ratio M, the design output with the highest matching degree is selected; the output design includes whole house style suggestions, space layout diagram, equipment suggestion list and construction budget estimate.

[0155] The construction of the smart home design model involves collecting rich historical matching cases to build input-output sample pairs, ensuring the diversity and representativeness of the model training data. User preference vectors Pp and natural apartment type fit values ​​are standardized and dimensionless normalized to ensure consistency in dimensions among different indicators, improving the scientific rigor and accuracy of matching calculations. Multiple weighted similarity matching methods combined with decision trees and deep learning models effectively integrate traditional algorithms with advanced artificial intelligence technologies. Cross-validation and early stopping mechanisms prevent overfitting, enhancing the model's generalization ability and robustness. By designing a preference matching ratio M to quantify the degree of fit between the apartment type and user needs, accurate filtering and ranking of recommendation results are achieved, greatly improving the personalization and matching accuracy of apartment type recommendations. The final output includes whole-house style suggestions, spatial layout diagrams, equipment lists, and budget estimates, meeting users' one-stop customization needs.

[0156] Experiments show that after adopting this method, the average subjective satisfaction of users with the recommended apartment layouts increased by 22.4%, and the matching rate of environmental comfort indicators (including lighting intensity, air exchange frequency, etc.) increased by more than 30%, demonstrating good application value and practicality.

[0157] Example 8

[0158] A smart home personalized design system, please refer to Figure 2 ,include,

[0159] The apartment layout data acquisition module is used to obtain residential structure diagrams, orientation and window opening information, and to construct apartment layout data groups;

[0160] The spatial analysis module is used to extract lighting and ventilation information to form the natural fit value for the i-th unit type. ;

[0161] The preference modeling module collects user member information, behavioral preferences, and budget expectations to generate a user preference vector Pp.

[0162] The design includes a large model building module for collecting historical matching case libraries and establishing input-output sample pairs for each case. The module standardizes the ratings in the user preference vector Pp to ensure the rating data is within a uniform dimensional range. Secondly, it calculates the natural fit value of the i-th apartment type. Dimensionless normalization is performed to ensure that the lighting and ventilation indicators match the dimensions preferred by users.

[0163] The model optimization and recommendation module is used to calculate the design preference matching ratio M, and preset the matching degree threshold X, and select several house types with a design preference matching ratio M higher than X as the final recommended solutions.

[0164] This system utilizes a floor plan data acquisition module with high-precision surveying technology to obtain residential structural diagrams, orientations, and window information, constructing detailed floor plan data sets to ensure the integrity and accuracy of basic data. The spatial analysis module, based on a lighting and ventilation physical model, extracts key environmental parameters to form the natural fit value for the i-th floor plan, enabling a quantitative evaluation of the floor plan's environmental performance. The preference modeling module comprehensively collects information on user family members, lifestyle preferences, and budget expectations, constructing a personalized user preference vector Pp to ensure that design schemes meet actual needs. The design model construction module combines floor plan data with user preference input, employing machine learning algorithms to train an intelligent design model, outputting diverse and highly matched home design schemes. The model optimization and recommendation module calculates the design preference matching ratio M, filters and ranks schemes with a matching degree higher than a preset threshold X, promoting accurate and personalized final recommendations.

[0165] The threshold is set to facilitate comparison. The size of the threshold depends on the amount of sample data and the number of bases set by those skilled in the art for each set of sample data; as long as it does not affect the ratio between the parameter and the quantized value, it is acceptable.

[0166] The above formulas are all derived from software simulation using a large amount of data and are selected to be close to the actual values. The coefficients in the formulas are set by those skilled in the art according to the actual situation. The above description is only a preferred embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any equivalent substitutions or changes made by those skilled in the art within the technical scope disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the protection scope of the present invention.

Claims

1. A method for personalized design of smart homes, characterized in that, Includes the following steps, Step 1: Obtain the apartment layout data of the target family and store it in the apartment layout data group. Extract features from the apartment layout data group and construct any two apartment layouts. and Structural similarity between And form the apartment space feature matrix S; Step 2: Based on the apartment type spatial feature matrix S, construct apartment type clusters using hierarchical clustering, perform structural similarity clustering analysis on all apartment types, and classify different types of residential space layouts; Step 3: Quantitatively analyze the environmental factors of the same type of apartment structure and calculate the spatial daylighting coefficient of the i-th apartment type. and ventilation index Construct the natural fit value for the i-th apartment type In addition, by combining the member composition and preference tags of the target family, a user preference vector Pp is constructed to form a design requirement data set; The spatial daylighting coefficient of the i-th apartment type and ventilation index The construction method is as follows: For the same type of apartment, extract window-related information for each room, including window location coordinates, orientation angle, and distribution of external obstructions. Based on geographic location information and shading models, and considering the orientation of room windows, the Radiance lighting simulation tool is used to calculate the sunshine duration Ts and maximum daylight intensity Ls for the corresponding apartment type, thus forming the spatial daylight coefficient for the i-th apartment type. ; ； In the formula, This indicates the weight of the sunshine duration Ts for the corresponding type of apartment. This indicates the weight of the maximum daylight intensity Ls for the corresponding apartment type. and All are constants, and ; It also extracts the door and window distribution information of each room, including the size, location, opening area, and connection with the outside or adjacent rooms. Based on the spatial layout, it analyzes the unobstructedness of the ventilation path and identifies the natural convection path, the angle between the air intake and exhaust, and the area A of each opening. Natural convection pathways include, In a north-south facing apartment, the south-facing windows bring in air while the north-facing windows exhaust air, creating the first cross ventilation path. Vertical convection created by the difference in the height of the opening This includes the upper window opening and the lower air inlet forming a thermal pressure-driven convection path, which allows hot air to rise and cold air to enter, forming a second vertical ventilation path; The connection between semi-open spaces such as balconies, terraces, and courtyards and interior doors and windows forms a third ventilation path; The interior atrium or central courtyard assists in the vertical circulation of airflow, creating a "chimney effect" that drives the air to flow up and down, forming a fourth ventilation path; Obtain the natural convection path, door and window convection angles, and opening areas to calculate the ventilation index of the i-th unit type. ; Where m represents the total number of identified natural convection paths. This represents the angle between the air intake and exhaust points of the j-th path. Let the effective opening area of ​​the j-th path be denoted as . This represents the path smoothness coefficient of the j-th path; in, This represents the total number of turns and bends along the j-th path. D represents the correction factor, where D=0.5 means that each turn reduces the traffic flow by 0.5 units. The natural fit value of the i-th apartment type The calculation formula is as follows: Combining the spatial lighting coefficient of the i-th unit and ventilation index Calculate the natural fit value of the i-th apartment type. , In the formula, and Let represent the daylighting coefficients of the i-th apartment type, respectively. and ventilation index The weight, , and These represent the maximum daylight intensity and maximum ventilation capacity observed in the same type of apartment, respectively, for normalization purposes; Step 4: Based on the design requirements data set, use the weighted similarity matching method to calculate the design preference matching ratio M between the natural fit value of the i-th apartment type and the user preference vector Pp, and output several successfully matched apartment types. Sort all apartment types, and preset the matching degree threshold X. Select several apartment types with a design preference matching ratio M higher than X as the final recommended scheme.

2. The smart home personalized design method according to claim 1, characterized in that, The intelligent surveying terminal is used to obtain the CAD drawings of the user's residential unit and the 3D scanning results. Intelligent mapping terminals include lidar equipment or panoramic scanners; Structured data of different residential units are extracted, including the number of rooms, layout area, opening orientation, window area, and floor orientation, to construct the feature set of the i-th unit type. The set of apartment type features of the j-th apartment type ; ； ； in, This represents the room distribution vector of the i-th apartment type. Let i represent the area vector of the i-th apartment type. Let represent the opening direction vector of the i-th apartment type. This represents the ratio of the window area to the total area of ​​the i-th apartment unit. This represents the floor height and orientation of the i-th apartment unit. This represents the room distribution vector of the j-th apartment type. Let the area vector of the j-th apartment type be represented. Let represent the opening direction vector of the j-th apartment type. This represents the ratio of the window area to the unit type j. Represents the floor height and orientation attributes of the j-th apartment type; sets the apartment type features of the i-th apartment type. The set of apartment type features of the j-th apartment type Store the apartment type data set H={F1, F2, ..., F n }; n represents the total number of apartment types; Calculate the similarity between any two apartment layouts using graphical analysis and spatial similarity formulas. and Structural similarity between ; in, Let represent the Euclidean distance between the sets of feature values ​​of the i-th and j-th apartment types. express and cosine similarity, and Represented as weights, , ; ; Based on structural similarity Construct the apartment layout spatial feature matrix S; In this matrix, the diagonal is 0, the similarity between matrices is constant, and the upper and lower triangles are symmetrical. Normalization and redundancy removal are then performed to obtain the processed apartment layout spatial feature matrix. .

3. The smart home personalized design method according to claim 2, characterized in that, The processed apartment space feature matrix Input a clustering analysis algorithm, construct apartment type clusters using hierarchical clustering, perform structural similarity clustering analysis on all apartment types, classify the apartment types within each cluster into classes with the same spatial structure, forming different residential space types, and construct an apartment type cluster set C; C = {C1, C2, ..., C6}. k }; where each cluster corresponds to a spatial structure type; and the structure class label of the k-th type of house is marked.

4. The smart home personalized design method according to claim 1, characterized in that, The steps for constructing the user preference vector Pp are as follows: Collect users' preferences for sunlight in their living environment. If users prefer ample sunlight, increase their "sunlight preference" score by 3 points; if they prefer soft lighting, adjust their "sunlight preference" score to a moderate level and increase it by 1-2 points. Collect users' preferences for sunlight. If a user prefers plenty of sunlight, increase their "sunshine preference" score by 3 points. If the user prefers moderate lighting, then the "sunshine preference" rating will be set to a medium score of 2 points. If the user prefers a soft or low-light environment, the "sunshine preference" score will be set to a low value of 1 point. Collect users' preferences for ventilation. If users prefer good ventilation, set the "ventilation preference" score to a high value of 3 points. If the user prefers moderate ventilation, the "ventilation preference" score will be set to a medium score of 2 points. If a user prefers a quiet environment with little ventilation, the "ventilation preference" score will be set to a low value of 1 point. By combining multiple user preferences, a user preference vector Pp is formed.

5. The smart home personalized design method according to claim 1, characterized in that, The construction method of the large-scale smart home design model is as follows: Collect a historical matching case library and create input-output sample pairs for each case; First, standardize the ratings in the user preference vector Pp to ensure the rating data is within a uniform dimensional range; second, standardize the natural fit value of the i-th apartment type. Dimensionless normalization is performed to ensure that the lighting and ventilation indicators match the dimensions preferred by users. Then, using a weighted similarity matching method, the design preference matching ratio M between the natural fit value of the i-th apartment type and the user preference vector Pp is calculated to output several successfully matched apartment types. The system employs decision tree or deep learning model algorithms to construct and train a large smart home design model. During training, cross-validation and early stopping mechanisms are introduced to prevent overfitting. After training, the model is applied to the test set, and the design preference matching ratio M is calculated to output several successfully matched house types. All apartment types are sorted, and a matching threshold X is preset. Several apartment types with a design preference matching ratio M higher than X are selected as the final recommended solutions.

6. The smart home personalized design method according to claim 5, characterized in that, The final recommended solution is obtained as follows: Based on the design preference matching ratio M, the design output with the highest matching degree is selected; the output design includes whole house style suggestions, space layout diagram, equipment suggestion list and construction budget estimate.

7. A smart home personalized design system, applied to the smart home personalized design method according to any one of claims 1-6, characterized in that, include, The apartment layout data acquisition module is used to obtain residential structure diagrams, orientation and window opening information, and to construct apartment layout data groups; The spatial analysis module is used to extract lighting and ventilation information to form the natural fit value for the i-th unit type. ; The preference modeling module collects user member information, behavioral preferences, and budget expectations to generate a user preference vector Pp. The design includes a large model building module for collecting historical matching case libraries and establishing input-output sample pairs for each case. The module standardizes the ratings in the user preference vector Pp to ensure the rating data is within a uniform dimensional range. Secondly, it calculates the natural fit value of the i-th apartment type. Dimensionless normalization is performed to ensure that the lighting and ventilation indicators match the dimensions preferred by users. The model optimization and recommendation module is used to calculate the design preference matching ratio M, and preset the matching degree threshold X, and select several house types with a design preference matching ratio M higher than X as the final recommended solutions.

Images