Apartment rental matching recommendation method and system based on multi-objective optimization

Abstract

The invention relates to an apartment rental matching recommendation method and system based on multi-objective optimization, and the method comprises the steps: obtaining user behavior data and market demand information, and generating an initial recommendation candidate set; a comprehensive score is calculated through a multi-dimensional matching algorithm, the weight is dynamically adjusted, and an optimization candidate set is generated; analyzing conflicting housing resources and performing secondary screening to obtain a final recommendation list for matching recommendation; according to the method, the accuracy of apartment rental matching recommendation can be improved by integrating multi-dimensional factors, the tenant satisfaction and the landlord rental efficiency are improved, and the limitation of single target guidance is solved.

Description

Technical Field

[0001] The present application relates to the fields of intelligent recommendation and optimization decision-making, and information technology, and in particular to a method and system for recommending apartment rental matching based on multi-objective optimization. Background Art

[0002] In modern society, demand for apartment rentals is growing, encompassing a diverse range of groups, including students, professionals, and families. With the acceleration of urbanization and increased population mobility, efficiently matching the needs of tenants and landlords has become a key issue, directly impacting the quality of rental services and market efficiency.

[0003] Most existing apartment rental recommendation systems are single-purpose, focusing on static matching based solely on rent or location. These systems typically rely on simple rules or pre-defined conditions to filter listings and generate a list of recommendations based on basic user input. However, in real-world rental scenarios, user needs are diverse and dynamically changing.

[0004] On the one hand, tenants' housing selection criteria are not limited to rent and location, but also involve multi-dimensional factors such as surrounding facilities and housing conditions. For example, young professionals may be more concerned about commuting convenience and living facilities, while family residents tend to consider educational resources and community environment. On the other hand, landlords' concerns have also expanded from simply renting out quickly to the long-term stability and satisfaction of tenants. Existing recommendation methods rarely take these multi-dimensional factors into consideration, resulting in deviations between recommendation results and user expectations, which may affect the tenant experience or reduce the landlord's rental efficiency. Therefore, a recommendation method that can integrate multi-dimensional factors and achieve two-way optimization is needed to improve the accuracy of apartment rental matching recommendations, thereby improving the overall level of rental services. Summary of the Invention

[0005] The main purpose of this application is to provide an apartment rental matching recommendation method and system based on multi-objective optimization to improve the accuracy of apartment rental matching recommendations.

[0006] To achieve the above objectives, an embodiment of the present invention provides a method for recommending apartment rental matching based on multi-objective optimization, the method comprising:

[0007] Obtaining historical user behavior data and market demand information, extracting user behavior features from the historical behavior data to obtain key behavior features, wherein the historical behavior data includes at least user browsing history, feedback scores, and interaction time distribution, and the market demand information includes at least apartment update frequency, supply-demand ratio, and regional popularity index;

[0008] Performing hierarchical cluster analysis on the market demand information to obtain a market supply and demand grouping model;

[0009] generating an initial apartment recommendation candidate set based on the key behavioral characteristics and the market supply and demand grouping model, wherein the initial apartment recommendation candidate set includes at least one candidate apartment;

[0010] performing a comprehensive score calculation for each candidate apartment in the initial recommended candidate set based on multiple comprehensive consideration dimensions of the candidate apartments, to obtain a comprehensive recommendation score for each candidate apartment in the initial recommended candidate set, wherein the multiple comprehensive consideration dimensions of the candidate apartments include at least rent compatibility, geographical convenience, completeness of facilities, and housing condition;

[0011] Optimizing the initial apartment recommendation candidates based on the comprehensive recommendation score and the user's current demand priority information to obtain an optimized apartment recommendation candidate set;

[0012] Identifying conflicting apartments in the optimized recommendation candidate set, wherein the conflicting apartments refer to apartments marked as highly preferred by multiple users within the same time period;

[0013] The optimized apartment recommendation candidate set is screened based on conflicting apartments to obtain a target apartment recommendation candidate set; and an apartment is pushed to the user according to the target apartment recommendation candidate set.

[0014] Accordingly, the embodiment of the present application also provides an apartment rental matching recommendation system based on multi-objective optimization, including

[0015] An acquisition module is configured to acquire user historical behavior data and market demand information, and extract user behavior features from the historical behavior data to obtain key behavior features. The historical behavior data includes at least user browsing history, feedback scores, and interaction time distribution. The market demand information includes at least apartment renewal frequency, supply-demand ratio, and regional popularity index.

[0016] A cluster analysis module, configured to perform hierarchical cluster analysis on the market demand information to obtain a market supply and demand grouping model;

[0017] a generating module, configured to generate an initial candidate set of recommended apartments based on the key behavioral characteristics and the market supply and demand grouping model, wherein the initial candidate set of recommended apartments includes at least one candidate apartment;

[0018] a scoring module configured to calculate a comprehensive score for each candidate apartment in the initial recommended candidate set based on multiple comprehensive consideration dimensions of the candidate apartments, thereby obtaining a comprehensive recommendation score for each candidate apartment in the initial recommended candidate set, wherein the multiple comprehensive consideration dimensions include at least rent compatibility, geographical convenience, completeness of facilities, and housing condition;

[0019] an optimization module, configured to optimize the initial apartment recommendation candidates based on the comprehensive recommendation score and the user's current demand priority information to obtain an optimized apartment recommendation candidate set;

[0020] an identification module, configured to identify conflicting apartments in the optimized recommendation candidate set, wherein the conflicting apartments refer to apartments marked as having high intention by multiple users within the same time period;

[0021] The recommendation module is used to perform apartment screening processing on the optimized apartment recommendation candidate set based on the conflicting apartments to obtain a target apartment recommendation candidate set; and perform apartment push processing on the user according to the target apartment recommendation candidate set.

[0022] In summary, the technical solution of the present application can accurately grasp user needs and market conditions by obtaining and processing user historical behavior data and market demand information. Through behavioral feature extraction and processing, key user behavior characteristics are mined, and market demand information is analyzed through hierarchical clustering to derive a market supply and demand grouping model. The two are combined to generate an initial recommendation candidate set, ensuring that the initial candidate apartments have a certain degree of accuracy and rationality. Then, a comprehensive recommendation score is calculated based on multiple consideration dimensions, and the candidate set is optimized based on the user's current demand priority to make the recommendation more in line with the user's current needs. By identifying and processing conflicting apartments, resource competition between users is avoided, and the target recommendation candidate set is screened and pushed to the user, thereby improving the accuracy and effectiveness of the recommendation. This solution overcomes the limitations of the single-goal orientation of existing rental recommendation systems. By comprehensively considering multiple factors such as rent, location, surrounding facilities, and housing conditions, combined with user behavior data and dynamic changes in market supply and demand, it improves the accuracy of apartment rental matching recommendations. It can not only provide users with more desired apartment recommendations, improve the efficiency and satisfaction of users in finding suitable apartments, but also improve the targeted rental of landlords, improve overall rental efficiency, and effectively improve the matching effect of the apartment rental market. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0024] Figure 1 Schematic diagram of a scenario for apartment rental matching recommendation based on multi-objective optimization in an embodiment of the present application;

[0025] Figure 2 A flowchart of the apartment rental matching recommendation method based on multi-objective optimization provided in an embodiment of the present application;

[0026] Figure 3 A schematic diagram of the process of extracting key behavioral features provided in an embodiment of the present application;

[0027] Figure 4 A schematic diagram of a scoring calculation method provided in an embodiment of the present application;

[0028] Figure 5 Another schematic diagram of a flow chart of a scoring calculation method provided in an embodiment of the present application;

[0029] Figure 6 A schematic diagram of the apartment screening process provided in an embodiment of the present application;

[0030] Figure 7 A schematic diagram of a process for calculating a conflict coefficient according to an embodiment of the present application;

[0031] Figure 8 A schematic diagram of the process of apartment push processing provided in an embodiment of the present application;

[0032] Figure 9 A schematic diagram of the structure of an apartment rental matching recommendation system based on multi-objective optimization provided in an embodiment of the present application;

[0033] Figure 10 A schematic diagram of the structure of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0034] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without making creative efforts are within the scope of protection of this application.

[0035] The embodiments of the present application provide a method and system for recommending apartment rental matching based on multi-objective optimization, which will be described in detail below.

[0036] In an embodiment of the present application, the apartment rental matching recommendation based on multi-objective optimization is a method that comprehensively considers multiple factors to achieve efficient apartment rental matching. It is based on the user's historical behavior data (such as browsing history, feedback ratings, interaction time distribution) and market demand information (apartment update frequency, supply-demand ratio, regional heat index). By extracting behavioral features from the user's historical behavior data to obtain key behavioral features, and performing hierarchical clustering analysis on the market demand information to obtain a market supply and demand grouping model, the two are combined to generate an initial recommendation candidate set. Then, a comprehensive score is calculated for the candidate apartments from multiple dimensions such as rent adaptability, geographical convenience, completeness of facilities and housing conditions, and the candidate set is optimized according to the user's current demand priority. Then, conflicting apartments are identified and processed, and finally the target recommendation candidate set is screened out and pushed to the user. This method aims to simultaneously improve tenant satisfaction and landlord rental efficiency, overcome the limitations of the single-goal orientation of traditional recommendation methods, and thus achieve more accurate and effective apartment rental matching recommendations in a complex apartment rental market environment.

[0037] like Figure 1 As shown, a scenario of an apartment rental matching recommendation method based on multi-objective optimization is provided. The apartment rental matching recommendation scenario based on multi-objective optimization mainly includes tenant information collection equipment, apartment information collection equipment, a multi-objective analysis and matching platform, and a recommendation information system. Data interaction is achieved between each device and platform through the network.

[0038] Tenant information collection equipment is used to comprehensively collect all types of tenant-related information. This data can be obtained through various means, such as the basic information provided by tenants when registering on rental platforms, including age, occupation, and family size. Age information can reflect a tenant's stage of life and specific needs. For example, younger tenants may prefer apartments close to entertainment venues or work, while older tenants may prioritize proximity to medical facilities and a quiet living environment. For example, tenants working in the financial industry may prefer apartments near a city's central business district for convenient commuting. Furthermore, this tenant information collection equipment collects tenants' browsing history, including details such as the apartment details they actively browse on the rental platform and the length of time they spend on search results. For example, if a tenant spends a significant amount of time browsing a page for an apartment with a gym, this may indicate a high level of interest in fitness facilities. The tenant information collection module also collects tenant feedback, such as their evaluation of previously rented apartments, their ratings of apartment cleanliness, and the landlord's service attitude. This information can reflect tenants' expectations and requirements for their living environment. The tenant information collection module organizes all collected information and sends it to the multi-objective analysis and matching system.

[0039] Apartment information collection equipment is responsible for collecting detailed information about apartments. This includes basic apartment attributes, such as location, down to the specific street address, and nearby transportation, such as walking distance to the nearest subway station or bus stop, and bus route coverage. Transportation accessibility is a key consideration for tenants. Rental information includes not only the monthly rent amount, but also payment methods (monthly, quarterly, or annual), as well as whether a deposit is required. Nearby amenities include detailed information on nearby supermarkets, shopping malls, hospitals, schools, parks, and other facilities. The availability of these amenities significantly impacts tenant convenience. Housing conditions are also key areas of focus, including apartment size, floor plan (one-, two-, or three-bedroom units), renovation level (fully furnished, minimally furnished, or bare), and orientation (sunlight exposure and lighting). This apartment information is continuously updated and transmitted to the multi-objective analysis and matching system.

[0040] The multi-objective analysis and matching platform can: extract user behavior characteristics from the historical behavior data to obtain key behavior characteristics; perform hierarchical clustering analysis on the market demand information to obtain a market supply and demand grouping model;

[0041] Based on the key behavioral characteristics and the market supply and demand grouping model, an initial apartment recommendation candidate set is generated, wherein the initial apartment recommendation candidate set includes at least one candidate apartment; a comprehensive score calculation is performed on each candidate apartment in the initial recommendation candidate set according to multiple comprehensive consideration dimensions of the candidate apartments to obtain a comprehensive recommendation score for each candidate apartment in the initial recommendation candidate set, wherein the multiple comprehensive consideration dimensions of the candidates include at least rental suitability, geographical convenience, completeness of facilities and housing conditions; based on the comprehensive recommendation score and the user's current demand priority information, the initial apartment recommendation candidates are optimized to obtain an optimized apartment recommendation candidate set; conflicting apartments in the optimized recommendation candidate set are identified, wherein the conflicting apartments refer to apartments marked as high-intent by multiple users in the same time period; and the optimized apartment recommendation candidate set is screened based on the conflicting apartments to obtain a target apartment recommendation candidate set.

[0042] The platform of the embodiment of the present application can conduct a multi-objective in-depth analysis of tenant and apartment information. The platform is built and optimized based on a large number of successful cases of tenant and apartment matching and market demand trends. First, the platform will conduct an in-depth analysis of tenant needs based on data such as the tenant's basic information and browsing history. For example, if the tenant is a young office worker and often browses pages of apartments located in bustling areas, with moderate rents and leisure facilities, the system will determine that the tenant pays more attention to the geographical location and surrounding entertainment facilities, and has a certain budget limit for rent. At the same time, the platform conducts a comprehensive evaluation of the apartment information, calculates the commuting time to the tenant's workplace based on the geographical location, calculates the richness score of the surrounding facilities, evaluates the living comfort score based on the housing conditions, and determines the cost-effectiveness score based on the rental information.

[0043] Then, the platform performs matching operations based on the principle of multi-objective optimization. If the tenant is sensitive to rent, the system will give priority to recommending apartments with lower rents, provided that the tenant's other needs are met (such as an acceptable commuting time and the presence of basic living facilities nearby). Assuming the tenant is a teacher, the system will be more inclined to recommend apartments that are close to the school and have a relatively quiet environment, while also considering whether the house type is suitable for living and working (for example, a relatively independent space may be needed for lesson preparation, etc.). When tenants have high requirements for quality of life, such as frequently browsing apartments with high-end decoration and high-end fitness facilities, the platform will increase the weight of house conditions and the completeness of surrounding high-end facilities during the matching process, and recommend apartments that better meet the tenant's expectations.

[0044] Based on the confirmed matching results, the multi-objective analysis and matching platform generates recommendations and sends them to the recommendation information push module. During this process, the platform also considers the interests of the landlord. For example, if a landlord wants an apartment to be rented quickly and to a long-term, stable tenant, the platform will comprehensively evaluate the tenant's stability factors, such as the tenant's career stability (tenants with long years of experience at large companies are relatively stable) and their residential history (previous rental periods, etc.), to try to recommend more suitable tenants to the landlord and improve the landlord's rental efficiency.

[0045] The recommendation information system is responsible for accurately delivering recommendations generated by the multi-objective analysis and matching platform to tenants. After the platform obtains a candidate set of target apartment recommendations, it can push information based on that set. For example, the system selects an appropriate push method based on the tenant's preferences. For example, for tenants who prefer to use mobile apps, recommended apartment information can be sent to them via app push notifications. This information includes images, detailed addresses, rent prices, key features (such as proximity to subway stations and nearby shopping malls), and landlord contact information. If the tenant prefers to receive information via email, the recommendations will be sent to them via email. Furthermore, the recommendation information push module adjusts push content based on tenant feedback. If a tenant is not interested in a particular apartment type (e.g., one that is too small), the push module will pass this feedback to the multi-objective analysis and matching system, which will then use this feedback to optimize subsequent matching and recommendation operations.

[0046] refer to Figure 2 , Figure 2 This is a flow chart of a method for recommending apartment rental matches based on multi-objective optimization provided in an embodiment of the present application. The method may be executed by a computer device, which may be a single computer device or a cluster of multiple computer devices. The computer device may be a terminal device or a server. The method for recommending apartment rental matches based on multi-objective optimization provided in an embodiment of the present application specifically includes:

[0047] S10: Obtain the user's historical behavior data and market demand information, extract the user's behavior features from the historical behavior data, and obtain key behavior features. The historical behavior data at least includes the user's browsing history, feedback score, and interaction time distribution. The market demand information at least includes the apartment update frequency, supply-demand ratio, and regional popularity index.

[0048] In this step, the user's historical behavior data refers to the collection of various related information generated by the user in the past interactions with the apartment rental platform.

[0049] Among them, browsing history is the historical trace of users viewing apartment information on the platform, including various information such as the geographical location, rent, and house structure of the viewed apartments; feedback score is the evaluation score given by users to apartments they have rented or viewed based on their own experience or expectations. This score reflects the user's satisfaction with different aspects of the apartment; interaction time distribution is the time statistics of users' interactive activities with the rental platform (such as browsing, querying, booking, etc.) in different time periods.

[0050] Market demand information is a collection of various data reflecting the overall state of the apartment rental market. Apartment renewal frequency refers to the number of newly available apartments or the frequency of updates to existing apartment listings within a specific timeframe. The supply-demand ratio refers to the ratio between tenant demand and the number of available apartments within a specific region or the entire market. The regional popularity index measures a specific area's popularity in the apartment rental market, taking into account factors such as regional foot traffic, economic development, and the availability of supporting facilities.

[0051] In one embodiment, an association rule mining method can be used to analyze browsing history. Taking the Apriori algorithm as an example, by analyzing the correlations between characteristics of apartments viewed by users, for example, frequently browsing low-priced apartments and apartments near subway stations, this indicates that low prices and convenient transportation are likely associated in the user's behavioral characteristics. For feedback scoring, sentiment analysis techniques can be used to convert users' textual reviews into quantified sentiment scores, with positive reviews receiving higher scores and negative reviews receiving lower scores. Through statistical analysis of these scores, user satisfaction with different apartment characteristics can be determined. For interaction time distribution, time series analysis methods can be used. For example, a 24-hour day can be divided into multiple time periods, and the frequency of user interactions within each time period can be counted. Time periods with higher interaction frequencies can be identified, which may indicate when users are most interested in apartment information. Through the aforementioned methods, key behavioral characteristics can be comprehensively derived. This method can comprehensively and deeply mine user behavioral patterns, providing a more accurate basis for subsequent precision recommendations and improving the matching of recommendation results with user needs.

[0052] In one embodiment, reference Figure 3 , step S10 may specifically include:

[0053] S101: segmenting the browsing records in the historical behavior data to obtain segmented time series.

[0054] In this step, browsing history represents the user's historical trajectory of browsing apartment listings on the apartment rental platform, including various attribute information of the apartments viewed at different points in time. Segmentation involves dividing continuous browsing history into different segments according to specific rules. The sequence of these segments is called a segmented time series. This segmentation facilitates more detailed analysis of user browsing behavior patterns over different time periods.

[0055] In one embodiment, segmentation can be performed based on fixed time intervals. For example, a 24-hour day can be divided into two-hour intervals, thus dividing the user's browsing history into 12 segments based on this time interval. For each segment, statistics are collected, including the number of times the user browsed apartments during that time period, the type of apartments viewed (e.g., by rental range, geographic location, etc.), and other information. This approach allows for simple and regular segmentation of browsing history, facilitating subsequent analysis of user browsing preferences within different time periods.

[0056] In one embodiment, segmentation can also be performed based on events. For example, each time a user logs into the rental platform is considered a start event, and the end event is the next login or when the user does not perform any operations for a long period of time (e.g., 30 minutes). The browsing history during this period is considered a segment. This approach better reflects the user's actual interaction process and can capture the changes in the user's browsing behavior during a complete platform usage. This approach can better reflect the user's browsing focus and trends during each use of the platform.

[0057] S102: Perform sliding window analysis on the segmented time series to extract the user's high-frequency access time periods.

[0058] In this embodiment, sliding window analysis is a data processing technique that analyzes the data within a fixed-size (time-length) window moving across a segmented time series. High-frequency access periods are defined as periods within these windows where users browse apartment information significantly more frequently than during other periods.

[0059] In one embodiment, a sliding window analysis with a fixed-size window can be used. For example, the window size is set to 1 hour, and starting from the beginning of the segmented time series, the window is moved by 15 minutes at a time. Within each window, metrics such as the number of apartment views and the frequency of viewing different apartment types are counted. When the number of views or frequency within a window exceeds a set threshold, the time period corresponding to that window is considered a high-frequency visit period. This analysis method can more accurately identify users' high-frequency browsing behavior within different time periods, providing a basis for understanding users' active time.

[0060] In one embodiment, a sliding window analysis with adaptive window size can also be used. This approach automatically adjusts the window size based on the data distribution. For example, a larger window size can be used in areas where browsing history data changes slowly, while a smaller window size can be used in areas where data changes dramatically. This approach can more flexibly adapt to different data characteristics and more accurately extract high-frequency access periods, thereby improving the accuracy of identifying high-frequency access periods, especially for browsing history data with large fluctuations.

[0061] S103: Perform sentiment analysis on the feedback scores to identify the user's positive preferences and negative rejection points.

[0062] Sentiment analysis, a natural language processing technique, analyzes the emotional tendencies underlying these ratings. Positive preferences refer to areas of satisfaction with apartments expressed in their ratings, while negative rejections represent areas of dissatisfaction.

[0063] In one embodiment, analysis can be performed based on a sentiment analysis method of a dictionary. Specifically, a sentiment dictionary containing positive words (such as "satisfied", "comfortable", "convenient", etc.) and negative words (such as "unsatisfied", "bad", "noisy", etc.) is constructed. Then, the text evaluation corresponding to the user's feedback score (if any) is analyzed word by word, and the frequency of occurrence of positive and negative words is counted. If the frequency of positive words is high, it indicates that there is a positive preference; if the frequency of negative words is high, there is a negative repulsion point. This analysis method can be simple and intuitive, and can quickly identify the basic emotional tendencies of users in text evaluations.

[0064] In one embodiment, sentiment analysis can also be performed using a machine learning algorithm, such as a support vector machine (SVM). First, a large number of feedback ratings with annotations (positive or negative sentiment) and their corresponding textual evaluation data are collected as a training set to train the SVM model. The user feedback ratings and textual evaluation data to be analyzed are then input into the trained model, which outputs a sentiment classification result, thereby determining positive preferences and negative rejection points. Machine learning methods have high accuracy and can handle relatively complex language expressions.

[0065] S104: Generate a user behavior feature map based on the high-frequency access time period, the positive preference, and the negative repulsion point to obtain key behavior features.

[0066] In this step, the user behavior feature map is a visual or structured representation of user behavior characteristics, such as high-frequency visit periods, positive preferences, and negative rejection points. This map can more intuitively display user behavior patterns, and the key behavioral features extracted from it can more accurately reflect user demand tendencies.

[0067] In one embodiment, high-frequency visit periods, positive preferences, and negative repulsion points can be used as different dimensions of the matrix. For example, rows represent different high-frequency visit periods, and columns represent different types of positive preferences and negative repulsion points (such as those classified by apartment rent, geographical location, facilities, etc.). In each cell of the matrix, relevant statistical information can be filled in, such as the frequency of attention to a certain positive preference type within a certain high-frequency visit period. Then, by analyzing this matrix, for example, calculating the weight and correlation of each row and column, key behavioral features can be extracted. This method can display the relationship between user behavioral features in a clear structure, facilitating the accurate extraction of key behavioral features.

[0068] In one embodiment, a graph structure can also be used to construct a user behavior feature map. Specifically, high-frequency access periods, positive preferences, and negative repulsion points are used as nodes in the graph, and edges are constructed based on the logical relationships between them (for example, a high-frequency access period may be associated with a positive preference). Key behavioral features are determined by analyzing the graph's topological structure, such as node degrees and path lengths. This approach can better capture the complex relationships between user behavior features and provide a more comprehensive perspective for extracting key behavioral features.

[0069] S20: Performing a hierarchical cluster analysis on the market demand information to obtain a market supply and demand grouping model.

[0070] In this step, hierarchical cluster analysis is a data analysis method that can gradually merge data objects into clusters at different levels based on the similarity between data objects (in this scenario, the similarity between various attributes in market demand information).

[0071] The market supply and demand grouping model is the result of a hierarchical cluster analysis of market demand information. It groups the market according to supply and demand characteristics. The market supply and demand conditions within each group are highly similar, while there are significant differences between groups. The market supply and demand grouping model is an abstract representation of the supply and demand conditions of the entire apartment rental market, derived through a hierarchical cluster analysis. It reflects the market supply and demand relationships and characteristics between and within different groups.

[0072] In one embodiment, a model can be constructed using a distance-based hierarchical clustering algorithm, such as the Euclidean distance algorithm. First, the various attributes in the market demand information, such as apartment renewal frequency, supply-demand ratio, and regional heat index, are considered as coordinates in a multidimensional space. For different regions or properties, their distances in this multidimensional space are calculated. Regions or properties with closer distances mean that they are more similar in terms of market supply and demand characteristics. Then, starting from each data point (i.e., each region or property), the closest points are merged into a cluster to form a hierarchical structure. In this process, the merging operation is repeated until a stopping condition is met, such as reaching a predetermined number of clusters or a threshold for intra-cluster distance. The resulting hierarchical structure is the market supply and demand grouping model. This clustering method can effectively classify complex market demand information, making the market supply and demand differences between different groups clear, which helps to make personalized recommendations for different market conditions in the subsequent recommendation process, improving the accuracy and pertinence of the recommendations.

[0073] In one embodiment, in order to improve the accuracy of the model, the market demand information can be subjected to a hierarchical clustering analysis to obtain multiple supply and demand groups, wherein each supply and demand group contains statistical results of real-time apartment update frequency, supply and demand ratio, and regional heat index; and a market supply and demand grouping model is generated based on the supply and demand groupings.

[0074] Hierarchical cluster analysis is a data analysis method that groups data objects based on similar characteristics. Specifically, data related to regions or apartment listings with similar market supply and demand characteristics can be grouped together.

[0075] The real-time apartment update frequency refers to the frequency of new or existing apartment information changes (such as rent adjustments, housing facility updates, etc.) on the market at the current moment; the supply-demand ratio is the ratio of the number of tenants' demand for apartments to the number of available apartments in a specific area or the entire market; the regional heat index is an indicator to measure the level of attention a region receives in the apartment rental market, which comprehensively considers multiple factors such as the region's economic development, population mobility, and supporting facilities.

[0076] In one embodiment, density-based hierarchical clustering is adopted. For example, each area or property is regarded as a point in the data space, and its coordinates are composed of the real-time apartment update frequency, supply-demand ratio and regional heat index. The DBSCAN algorithm is based on the density of points. If the number of points within a certain radius around an area or property (i.e., areas or properties with similar update frequency, supply-demand ratio and heat index) reaches a certain threshold, these points are divided into a preliminary cluster. Then, these preliminary clusters are further hierarchically constructed, for example, density-connected clusters are merged into larger clusters to obtain multiple supply and demand groups. This method can effectively discover the data distribution of different density areas in the data set, has a good effect on processing the supply and demand data distribution of complex shapes, and can accurately group according to the intrinsic characteristics of market demand information.

[0077] In one embodiment, agglomerative hierarchical clustering in the hierarchical clustering algorithm can also be used. Initially, each area or property is regarded as a separate group. Then, the distance between each two groups is calculated. The distance here can be defined by comprehensively considering the real-time apartment update frequency, the supply-demand ratio and the difference in regional heat index, such as using Euclidean distance or Manhattan distance. The two groups with the closest distance are selected for merging, and this process is repeated until the stopping condition is met, such as reaching a predetermined number of clusters or a threshold value of the distance between groups. In this way, multiple supply and demand groupings are formed. The use of agglomerative hierarchical clustering does not require pre-specifying the number of clusters, and can naturally form a hierarchical clustering result, which is more effective for exploring the inherent hierarchical structure of market demand information.

[0078] In one embodiment, a graph-based market supply and demand grouping model can be constructed. Specifically, each supply and demand group is treated as a node in the graph, and edges between nodes represent the relationships between different groups. For example, edge weights can be determined based on the similarity or complementarity of supply and demand characteristics between groups. If two groups have significant similarities in apartment renewal frequency, supply-demand ratio, and regional popularity index, then the edge weight between them will be higher. If one group has an oversupply of housing while another group has strong demand, and there is a correlation based on geographic location or other factors, this complementary relationship can also be represented by an edge. This approach can intuitively demonstrate the complex relationships between different supply and demand groups, facilitating comprehensive consideration of these relationships during subsequent apartment recommendations. For example, if housing supply is tight within a group, the availability of housing in other closely related groups can be considered.

[0079] In one embodiment, a market supply and demand grouping model can also be constructed in matrix form. For example, a matrix is ​​created with rows and columns corresponding to different supply and demand groups. The elements in the matrix represent the interactions or characteristic differences between different groups. For example, the matrix can be populated with information such as the differences in supply and demand ratios between different groups, the correlation coefficient of regional heat index, etc. This method can conveniently analyze and process the relationships between different groups through matrix operations, such as calculating the similarity matrix between groups, providing a quantitative basis for further market analysis and apartment recommendations.

[0080] S30: Based on the key behavioral characteristics and the market supply and demand grouping model, an initial apartment recommendation candidate set is generated, where the initial apartment recommendation candidate set includes at least one candidate apartment.

[0081] In the embodiment of the present application, the initial apartment recommendation candidate set is a set of apartments that may meet the user's needs and are preliminarily screened from the apartment library based on the user's key behavioral characteristics and the market supply and demand grouping model.

[0082] In one embodiment, the user's demand weights for different apartment features can be determined based on key behavioral characteristics. For example, if the user's browsing history shows that they frequently browse low-rent apartments and their feedback score shows a high sensitivity to rent, then the demand weight for rent adaptability will be higher. Then, based on the market supply and demand grouping model, areas or properties that meet the market supply and demand conditions are screened out. For example, in areas with a high supply-demand ratio, those properties with high cost-effectiveness may be given priority. For each property that meets the market supply and demand conditions, its matching degree with the user's needs is calculated according to the user's demand weight. Properties with a matching degree that reaches a certain threshold are included in the initial apartment recommendation candidate set. This method can comprehensively consider user needs and market conditions to generate an initial recommendation candidate set that not only meets the user's potential expectations but also has market feasibility, providing a reasonable basis for subsequent accurate recommendations, reducing the scope of unnecessary property screening, and improving recommendation efficiency.

[0083] S40: Based on the multiple comprehensive consideration dimensions of the candidate apartments, a comprehensive score calculation is performed on each candidate apartment in the initial recommendation candidate set to obtain a comprehensive recommendation score for each candidate apartment in the initial recommendation candidate set, wherein the multiple comprehensive consideration dimensions of the candidate apartments at least include rental suitability, geographical convenience, completeness of facilities and housing conditions.

[0084] In this step, rental suitability refers to the degree to which the rental of the candidate apartment matches the user's budget and the average market rent.

[0085] Geographical convenience refers to the locational advantage of a candidate apartment relative to the user's workplace, frequent activities, or transportation hub, including distance and accessibility. Amenities refer to the availability of various amenities (such as shopping malls, hospitals, schools, and entertainment venues) around the candidate apartment.

[0086] Housing conditions cover the internal physical properties of the candidate apartment, including size, layout, decoration quality and other aspects.

[0087] The comprehensive recommendation score is a score calculated based on comprehensive considerations of the apartment through a specific calculation method to measure the overall degree to which the candidate apartment meets user needs.

[0088] In one embodiment, a weighted approach can be used to calculate the rating. First, the weights of the comprehensive consideration dimensions for each apartment are determined. For example, for young professionals, locational convenience may have a higher weight, while for families, the housing conditions of the apartment and the completeness of surrounding amenities may have higher weights. Then, for each candidate apartment, scores are calculated for each of the following dimensions: rent compatibility, locational convenience, completeness of amenities, and housing condition. The rent compatibility score can be determined by comparing the apartment's rent to the user's budget range. For example, if the rent is within the budget range and close to the lower limit, the score will be higher. The locational convenience score can be calculated based on the distance from the user's workplace and transportation accessibility. The closer the distance and the more convenient the transportation, the higher the score. The completeness of amenities score can be quantified based on the type and number of surrounding amenities. For example, the more types of surrounding amenities, the higher the score. The housing condition score can be comprehensively evaluated based on factors such as area, reasonableness of the apartment layout, and quality of decoration. Finally, the scores of each dimension are multiplied by their corresponding weights and summed to obtain a comprehensive recommendation score for each candidate apartment. This method can comprehensively and quantitatively evaluate each candidate apartment, making the recommendation results more objective and comparable, helping to sort candidate apartments according to the comprehensive score, and providing users with a more intuitive selection reference.

[0089] In one embodiment, the rental suitability, geographical convenience, completeness of facilities and housing conditions of the candidate apartments are weighted to obtain a comprehensive recommendation score for each apartment in the initial recommendation candidate set.

[0090] Specifically, a weighted calculation can be performed based on weights determined through data mining. First, a large amount of data on completed apartment rental transactions and corresponding user evaluation data can be collected. By analyzing this data, the correlation between rent compatibility, location, amenities, and housing conditions, and the user's final satisfaction level can be calculated. Dimensions with higher correlations are assigned higher weights. For example, if location is found to have the highest correlation with user satisfaction, it will be given a higher weight. Then, the scores for each dimension are calculated and weighted together as described above to obtain a comprehensive recommendation score. This method of determining weights based on actual data better reflects the actual market situation and user needs, making the comprehensive recommendation score more objective.

[0091] S50: Optimizing the initial apartment recommendation candidates based on the comprehensive recommendation score and the user's current demand priority information to obtain an optimized apartment recommendation candidate set.

[0092] In this step, the user's current demand priority information refers to the ranking information of the importance the user places on different apartment features (such as rent, geographical location, facilities, etc.) at the current moment.

[0093] The optimized apartment recommendation candidate set is a set of apartments that better meets the user's current needs, obtained by adjusting the initial apartment recommendation candidate set based on the comprehensive recommendation score and the user's current demand priority information.

[0094] In one embodiment, the user's current demand priority information can be converted into a quantifiable adjustment factor. For example, if the user currently increases the priority of geographical convenience, the adjustment factor corresponding to geographical convenience will be greater than 1. Then, for each apartment in the initial apartment recommendation candidate set, the score of each dimension is recalculated using the adjustment factor based on its comprehensive recommendation score. For example, the score of geographical convenience is multiplied by the corresponding adjustment factor and then the comprehensive recommendation score is recalculated. Finally, based on the recalculated comprehensive recommendation score, the top-ranked apartments are screened out to form an optimized apartment recommendation candidate set. This method can respond to the dynamic changes of user needs in a timely manner, make the recommendation results more in line with the user's current actual needs, and improve the user's satisfaction with the recommendation results.

[0095] S60: Identify conflicting apartments in the optimized recommendation candidate set, where the conflicting apartments refer to apartments marked as highly preferred by multiple users within the same time period.

[0096] In this embodiment of the application, conflicting apartments are apartments in the optimized recommendation candidate set that are marked as having high rental intention by multiple users within the same time period. This situation may cause multiple users to compete for the same apartment, thus affecting the effectiveness and fairness of the recommendation.

[0097] In one embodiment, a tag counter can be set for each apartment. When a user marks an apartment as having high preference, the corresponding counter is incremented by 1. Within the optimized recommendation candidate set, all apartments are traversed and their tag counter values ​​are checked. If the counter value is greater than 1, indicating that the apartment has been marked as having high preference by multiple users within the same time period, the apartment is identified as a conflicting apartment. This technical implementation method allows for simple and efficient identification of conflicting apartments, providing an accurate basis for subsequent screening and improving the stability and reliability of the recommendation system.

[0098] S70: performing apartment screening processing on the optimized apartment recommendation candidate set based on the conflicting apartments to obtain a target apartment recommendation candidate set; and performing apartment push processing on the user according to the target apartment recommendation candidate set.

[0099] In this step, the target apartment recommendation candidate set is the final apartment set recommended to the user after filtering out conflicting apartments based on the optimized apartment recommendation candidate set.

[0100] In one embodiment, for the identified conflicting apartments, a priority is set for them based on other relevant factors. For example, if the landlord of one of the conflicting apartments has a higher reputation, or if it is more scarce in the market (such as a unique geographical location or special housing facilities), then this apartment will have a higher priority. For non-conflicting apartments, their original order in the optimized apartment recommendation candidate set is maintained. Then, according to the set priority and original order, a certain number of apartments are screened from the optimized apartment recommendation candidate set to form the target apartment recommendation candidate set. This method can retain apartments that meet user needs to the greatest extent while handling conflicting apartments, ensuring that the recommendation results can both meet user needs and reasonably resolve apartment conflicts, thereby improving the rationality and effectiveness of the recommendations.

[0101] After obtaining the target apartment recommendation candidate set, the embodiment of the present application can perform apartment push processing on the user based on the candidate set. Push processing can be implemented in a variety of ways. For example, when the user logs in to the client of the rental platform (mobile APP or web page), the apartment information in the target apartment recommendation candidate set is displayed to the user in a certain format (such as picture display, text description of rent, geographical location, surrounding facilities and other important information). The order of push can be arranged according to the comprehensive recommendation score of the apartment in the target apartment recommendation candidate set or other relevant factors (such as the degree of match with user preferences). The technical effect of this push processing is that it can recommend apartments that meet user needs to users in a timely and accurate manner, improve the efficiency of users in obtaining suitable apartments, and enhance user experience.

[0102] In one embodiment, in order to accurately calculate the recommendation score of the candidate house or apartment in the candidate set, reference Figure 4 , can be calculated as follows:

[0103] S4011: Calculate individual scores for rental suitability, geographical convenience, completeness of facilities, and housing conditions for each candidate apartment in the initial recommended candidate set.

[0104] In this step, the calculation of a single score for rental suitability aims to quantify the suitability of the candidate apartment's rent with the user's budget and the market rental level.

[0105] The single score of geographical convenience is a quantitative assessment of the degree of geographical advantage of the apartment relative to the user's specific needs (such as workplace, frequently visited commercial center, etc.).

[0106] The "Amenities" score is a numerical representation of the availability of amenities around the apartment (such as supermarkets, hospitals, schools, and recreational facilities). The "Housing Condition" score is a quantitative evaluation of the apartment's physical attributes, including size, floor plan, and finish quality.

[0107] In one embodiment, a formula can be established for the individual scoring of rental suitability. For example, if the apartment rent is within the user's budget and close to the lower limit, the score is higher. The score can be assigned between 0 and 100 based on the proximity to the lower limit, with higher scores being assigned the closer to the lower limit. For the individual scoring of geographical convenience, if the apartment is within a 10-minute walk from the user's workplace, the score is 80-100 points; if it is 10-20 minutes, the score is 60-80 points, and so on. For the individual scoring of facility completeness, a base score can be set based on the type and number of surrounding facilities, with additional points added for each additional important facility (such as a hospital or large supermarket). For the individual scoring of housing conditions, for example, different score ranges can be assigned based on whether the area meets the user's needs, whether the apartment layout is reasonable, and the quality of the decoration, and then a comprehensive calculation is performed. This calculation method standardizes the calculation of individual scores through clear rules, making it easy to operate and the results interpretable.

[0108] In one embodiment, a large amount of data on rented apartments can be collected, including rent, location, surrounding amenities, property conditions, and tenant satisfaction scores. Using machine learning algorithms, such as regression analysis, a relationship model is established between rent compatibility, location convenience, amenities, property conditions, and tenant satisfaction. The model then incorporates the relevant data for each apartment in the initial set of recommended candidates to generate a corresponding individual score. This approach leverages large amounts of data to uncover potential relationships between various factors and user satisfaction, ensuring that individual scores are more tailored to actual user needs.

[0109] S4012: Utilize the nonlinear mapping layer in the multidimensional matching algorithm to normalize the individual scores to obtain standardized individual scores.

[0110] In this step, the multidimensional matching algorithm comprehensively considers multiple dimensions to perform matching calculations. A nonlinear mapping layer is a component of this algorithm. Its function is to perform a nonlinear transformation on the input individual scores, thereby converting individual scores with different value ranges and distributions into standardized individual scores under a unified standard. The purpose of normalization is to make the individual scores comparable and eliminate the impact of differences in the original data's dimensions, value range, and other factors on the subsequent weighted calculations.

[0111] In one embodiment, a Sigmoid function can be used for nonlinear mapping normalization. Substituting the individual scores into this function yields a value between 0 and 1, which is the standardized individual score. For example, an apartment's locational convenience score of 80 points can be converted to a value between 0 and 1, such as 0.95, using the Sigmoid function. Because the Sigmoid function has smooth nonlinear characteristics, it can map input values ​​of varying ranges to a relatively stable range. It is commonly used in data processing and is easy to implement and understand.

[0112] S4013: Based on the standardized individual scores, the comprehensive recommendation score is calculated using the weighted fusion layer in the multidimensional matching algorithm; the weighted fusion layer dynamically adjusts the weight of each individual score according to the priority of user needs.

[0113] In this embodiment, the weighted fusion layer is the component of the multi-dimensional matching algorithm responsible for weighted summing of standardized individual scores to produce a comprehensive recommendation score. It dynamically adjusts the weights of individual scores based on user priority. This means it can adjust the weight of each factor in the comprehensive score in real time based on the user's current emphasis on factors such as rent compatibility, location convenience, complete amenities, and housing condition.

[0114] In one embodiment, some rules related to changes in user demand priorities can be pre-set. For example, if a user frequently searches for low-priced apartments on the platform, it indicates that the demand priority for rent compatibility has increased. At this time, the weighted fusion layer will increase the weight of rent compatibility by a certain percentage according to the preset rules, such as from the original 0.2 to 0.3, while adjusting other weights accordingly (the total remains 1). Then, the standardized individual scores are multiplied by the adjusted weights and the sum is calculated to obtain the comprehensive recommendation score.

[0115] In one embodiment, a neural network model can be constructed to implement a weighted fusion layer. The standardized individual scores and user demand priority information are used as inputs to the neural network, and the neural network learns the adjustment mode of the individual score weights under different user demand priorities through training. For example, the input is the standardized rent suitability, geographical convenience, facility completeness, housing condition score and user demand priority vector (such as a vector indicating the degree of importance attached to rent suitability, geographical convenience, etc.), and the output is the adjusted individual score weights. The standardized individual scores are then multiplied by these weights and summed to obtain a comprehensive recommendation score. This method can automatically learn the complex relationship between user demand priority and weight adjustment, and has better adaptability to various changes in user demand.

[0116] In one embodiment, in order to improve the accuracy of apartment matching and recommendation, the rental suitability, geographical convenience, completeness of facilities and housing conditions of the candidate apartments are weighted to obtain a comprehensive recommendation score for each apartment in the initial recommendation candidate set; at this time, the reference Figure 5 , step S50 can be specifically implemented as follows:

[0117] S501: Configuring a dynamic weight adjustment mechanism based on the comprehensive recommendation score.

[0118] In this embodiment, the dynamic weight adjustment mechanism is a rule or model that dynamically adjusts the weights of the various apartment considerations used to calculate the comprehensive recommendation score based on changes in user priority information. The comprehensive recommendation score reflects the overall degree to which the candidate apartment meets user requirements in its initial state, and the dynamic weight adjustment mechanism aims to further optimize this scoring system based on the dynamic changes in user needs.

[0119] In one embodiment, pre-defined rules can be used to adjust weights. For example, if a user frequently changes their budget within a specific time period, the sensitivity of the rental adaptability weight adjustment can be increased; if the user's work location changes, the adjustment amplitude of the geographical convenience weight can be increased. When changes in user demand priority information are detected, the weights of each dimension are adjusted according to these pre-defined rules. This approach is simple and intuitive, easy to understand and implement, and can quickly adjust weights based on common scenarios of changing user needs.

[0120] In one embodiment, a machine learning model can be used to construct a dynamic weight adjustment mechanism. For example, a neural network model can be used to take the comprehensive recommendation score, user demand priority information, and other relevant auxiliary information (such as changes in user browsing behavior, recent search keywords, etc.) as input, and output the adjusted weights of the comprehensive consideration dimensions of each apartment. The neural network is trained with a large amount of sample data so that it can accurately adjust the weights according to different input conditions. This method has strong adaptability and flexibility, can handle complex changes in user needs, and improve the accuracy of weight adjustment.

[0121] S502: Adjust the weight of the comprehensive consideration dimension of each apartment according to the demand priority information and the dynamic weight adjustment mechanism.

[0122] In this step, demand priority information clarifies the user's current ranking or preference for apartment factors such as rent compatibility, location, amenities, and housing conditions. The dynamic weight adjustment mechanism provides a specific method for adjusting weights based on demand priority information.

[0123] In one embodiment, the weights can be adjusted using a linear adjustment method. Assuming that, based on the demand priority information, the priority of geographical convenience is determined to be increased by a certain proportion, according to the linear rule in the dynamic weight adjustment mechanism, the original weight of geographical convenience is multiplied by an adjustment coefficient greater than 1 (this coefficient is determined according to the degree of priority increase), and the weights of other dimensions are correspondingly reduced proportionally to ensure that the sum of the weights of all dimensions is 1. For example, if the original weight of geographical convenience is 0.3, and its priority is determined to be increased by 20% based on the demand priority information, the adjusted weight is 0.3×(1+0.2)=0.36, and the weights of other dimensions are redistributed proportionally. This method is simple to calculate, can intuitively adjust the weights according to the demand priority, and maintain the relative stability of the weight system.

[0124] In one embodiment, an adjustment method based on function mapping can also be used. A mapping function is constructed based on demand priority information. For example, for rent adaptability, if the demand priority decreases, the mapping function can be a decreasing power function. The original weight of the rent adaptability is adjusted through this mapping function according to the dynamic weight adjustment mechanism. The same applies to other dimensions. This method can more flexibly adjust weights based on demand priority information and is particularly suitable for nonlinear demand changes.

[0125] In one embodiment, to improve the accuracy of apartment recommendations, the weights of the comprehensive consideration dimensions of each apartment can be adjusted as follows:

[0126] Step A: Acquire the user's real-time demand priority information, and parse the demand priority information to obtain semantic description information of the demand priority information.

[0127] In this step, users' real-time demand priority information refers to the dynamic information about their current ranking or preference for different apartment considerations (such as rent compatibility, location, amenities, and housing conditions). Parsing is the process of converting this demand priority information into meaningful semantic descriptions.

[0128] First, real-time demand priority information can be obtained from a variety of sources. For example, it can come from user interactions on the rental platform, such as resetting the rental range or adjusting the location filter during search. It can also be based on analysis of users' recent browsing behavior patterns. The goal of parsing is to transform this raw operational or behavioral data into understandable semantic content.

[0129] In one embodiment, a set of rules can be pre-defined. For example, if a user narrows the rental range and increases the upper limit of the rental on the platform, the rules can be interpreted as "the user's priority for rental adaptability has increased, and they prefer apartments with higher rents to meet other potential needs (such as better housing conditions or complete facilities)." If the user limits the search scope to apartments within a certain distance from a specific location (such as the workplace), it can be interpreted as "the priority of the need for geographical convenience has significantly increased." This method can be simple and direct, and can quickly parse common user operations according to predefined rules to obtain semantic description information.

[0130] In one embodiment, natural language processing technology can also be used. When the user expresses the priority of his or her needs in natural language form (such as entering text in the demand description box of the platform), the sentence entered by the user is parsed using technologies such as lexical analysis and syntactic analysis. For example, the user enters "I hope to give priority to apartments that are close to subway stations and have rents below 3,000 yuan." Lexical analysis identifies that "close to subway stations" is related to geographical convenience, and "rents below 3,000 yuan" is related to rent adaptability. Syntactic analysis is then used to determine the priority relationship between the two, thereby obtaining semantic description information. The technical effect of this technical implementation method is that it can handle complex needs expressed by users in natural language, and improve the accuracy and flexibility of parsing.

[0131] Step B: Quantitatively model the semantic description information to generate a priority weight vector.

[0132] In this step, semantic description information is the content with clear meaning obtained after parsing the user demand priority information. Quantitative modeling converts this semantic information into a quantitative form that can be used for calculation. The generated priority weight vector is a vector that can reflect the weight of the comprehensive consideration dimensions of each apartment.

[0133] For quantitative modeling, the weight of each consideration dimension needs to be determined based on the semantic description information. For example, if the semantic description information indicates that users place great importance on location, place moderate emphasis on rental suitability, and place relatively little emphasis on amenities and housing conditions, then this difference in importance needs to be converted into specific numerical values ​​during quantitative modeling.

[0134] In one embodiment, quantitative modeling is performed based on the analytic hierarchy process (AHP). First, a hierarchical model is constructed, and the target layer is set to generate a priority weight vector. The criterion layer is the four comprehensive consideration dimensions of the apartment, namely, rent adaptability, geographical convenience, completeness of facilities, and housing conditions. The solution layer is the different demand priority solutions determined according to the semantic description information. Then, by constructing a judgment matrix, the relative importance of each element in the criterion layer is compared. For example, the importance of geographical convenience relative to rent adaptability is judged according to the semantic description information, and the corresponding numerical value is filled in the judgment matrix (usually using a 1-9 scaling method). Finally, the priority weight vector is obtained by calculating the eigenvector of the judgment matrix and performing a consistency test. The technical effect of this technical implementation method is that it can systematically consider the relative importance between each consideration dimension, obtain a reasonable weight vector through a scientific calculation method, and perform a consistency test to ensure the reliability of the results.

[0135] Step C: Dynamically adjust the weight ratios of the rental suitability, geographical convenience, facility completeness, and housing condition based on the priority weight vector.

[0136] In this step, the priority weight vector clarifies the weight relationship of each apartment's comprehensive consideration dimensions under the current user demand priority. Dynamic adjustment is to change the original weight ratio of rental adaptability, geographical convenience, completeness of facilities, and housing conditions based on this weight vector.

[0137] In one embodiment, the originally set weights for rent compatibility, geographical convenience, facility completeness, and housing condition can be directly replaced with the corresponding values ​​in the priority weight vector. For example, if the original weight of rent compatibility is 0.25, geographical convenience is 0.3, facility completeness is 0.2, and housing condition is 0.25, if the weight of rent compatibility is changed to 0.3, geographical convenience is changed to 0.4, facility completeness is 0.15, and housing condition is 0.15 according to the priority weight vector, then these new weights are directly applied to subsequent calculations. This method is simple and direct to operate, and can quickly adjust the weight ratio according to the priority weight vector.

[0138] In one embodiment, a proportional adjustment method can also be used. The proportional relationship between the priority weight vector and the original weight vector is calculated, and then the original weight is adjusted according to this proportional relationship. For example, the original weight vector is , the priority weight vector is , and the proportional relationship is calculated to be . The original weights are then multiplied by the corresponding proportional coefficients to obtain the adjusted weights. This method can maintain a certain correlation with the original weights when adjusting the weight ratio, avoid overly abrupt weight adjustments, and better meet the requirement for smooth weight adjustment.

[0139] S503: Based on the adjusted weights, the rental suitability, geographical convenience, completeness of facilities and housing conditions of the candidate apartments are weighted to obtain a comprehensive recommendation score for each apartment in the initial recommendation candidate set.

[0140] In this step, the weights adjusted in the previous step can more accurately reflect the user's current demand priorities. The dimensions of the candidate apartments are weighted again to obtain a comprehensive recommendation score that better meets the user's current needs.

[0141] In one embodiment, weighted calculation can be performed directly according to the adjusted weights. First, for each candidate apartment, its scores in dimensions such as rent compatibility, geographical convenience, completeness of facilities and housing conditions are recalculated (the calculation method can be the same as before). Then, the scores of each dimension are multiplied by the adjusted weights and summed to obtain a new comprehensive recommendation score. For example, the rent compatibility score of a candidate apartment is 80 points, the adjusted weight is 0.2, the geographical convenience score is 70 points, the adjusted weight is 0.3, the facility completeness score is 60 points, the adjusted weight is 0.2, the housing condition score is 75 points, the adjusted weight is 0.3, then the comprehensive recommendation score = 80×0.2+70×0.3+60×0.2+75×0.3=71.5 points. This method can quickly obtain a new comprehensive recommendation score based on the adjusted weights, and intuitively reflect the impact of weight adjustment on the comprehensive recommendation score.

[0142] In one embodiment, the comprehensive recommendation score can also be calculated using an incremental update method. If the original comprehensive recommendation score has been calculated before and the adjustment amount of each dimension weight is known, a new comprehensive recommendation score can be obtained by incremental calculation. For example, assuming that the original comprehensive recommendation score is 70 points, the rental adaptability weight is increased by 0.05, and its score is 80 points, the geographical convenience weight is reduced by 0.05, and its score is 70 points, the facility completeness and housing condition weights remain unchanged, and their scores are 60 points and 75 points respectively, then the new comprehensive recommendation score = 70 + 80 × 0.05 - 70 × 0.05 = 70 + 4 - 3.5 = 70.5 points. This method is to perform incremental updates based on the existing calculation results, which reduces the amount of calculation and improves the calculation efficiency.

[0143] In one embodiment, reference Figure 6 , step S60 can be implemented in the following ways:

[0144] S601: Mapping the conflicting apartments to an apartment allocation grid model, and calculating the conflict coefficient of the conflicting apartments based on the user's time preference and apartment reservation status.

[0145] Among them, conflicting apartments refer to apartments that are marked as high-intent by multiple users in the same time period.

[0146] Among them, the apartment allocation grid model is a model structure used to analyze and process the resource allocation situation of apartments. It can present various status information of apartments in an intuitive and easy-to-calculate way.

[0147] Among them, the user's time preference refers to the user's tendency towards time-related factors such as check-in time and rental duration when renting an apartment.

[0148] Among them, the apartment reservation status refers to relevant information such as whether the apartment is currently booked, the booking time period, and the remaining time period available for rental.

[0149] Among them, the conflict coefficient is a quantitative indicator used to measure the degree of conflict in a conflict apartment under multi-user competition.

[0150] In one embodiment, a two-dimensional apartment allocation grid model can be constructed. Within this model, one dimension represents time (which can be divided into time units such as days, weeks, and months), and another dimension represents different apartment attributes (such as geographic location and apartment type) or directly represents individual apartments. Conflicting apartments are mapped to corresponding locations in this grid model based on their corresponding time attributes and their own attributes. For example, if a conflicting apartment is located in a specific location and is expected to be available for rent in a certain month, the apartment is marked at the corresponding location in the grid model. Regarding user time preferences, suppose user A prefers to move in immediately and have a lease term of 6 months, while user B prefers to move in one month later and have a lease term of 3 months. Regarding apartment reservation status, if an apartment is partially booked, the booked time interval is marked. When calculating the conflict coefficient, the degree of time overlap, the urgency of the user's time preference, and the degree of apartment reservation tension can be comprehensively considered. For example, the proportion of time overlap to the user's desired rental time period is calculated. Urgent user time preferences (e.g., an urgent need for immediate move-in) are given a higher weight, as are tight apartment reservations (e.g., a short available time period). The conflict coefficient is then calculated using a weighted formula. The technical effect of this technical implementation method is to clearly present the relationship between apartments, time and user needs through a two-dimensional grid model, and to calculate the conflict coefficient by considering various factors more comprehensively.

[0151] In one embodiment, a graph-based apartment allocation grid model can also be used. Apartments, users, and time are considered nodes in the graph, and edges are constructed based on the relationships between them. For example, conflicting apartments are connected to users with high interest in them by edges, and the different rental time periods of apartments are also connected to apartment nodes by edges. The weight of the user node is set based on the user's time preference. For example, the more urgent the time preference, the higher the weight. The weight of the apartment node in different time periods is set based on the apartment's reservation status. The shorter the rental time period, the higher the weight. The conflict coefficient is then calculated using a graph algorithm (such as a variation of the shortest path algorithm). This coefficient reflects the degree of conflict between the conflicting apartments under this complex graph structure relationship. This method can well handle complex relationship structures and more flexibly calculate the conflict coefficient based on various factors. It is particularly suitable for situations where multiple factors influence each other.

[0152] In one embodiment, in order to improve the efficiency and accuracy of conflict coefficient calculation, reference Figure 7 , step S601 can be implemented by the following process:

[0153] S6011: Obtain the apartment marking status and time preference of each candidate apartment in the optimized apartment recommendation candidate set, and query the current reservation status and historical rental period of the candidate apartment.

[0154] In this step, the optimized set of apartment recommendation candidates is a collection of apartments that have been screened and optimized to better meet user needs. Apartment tagging status refers to information about apartments marked by users in the recommendation system, such as whether they are marked as interesting or highly desirable. Time preference refers to the user's preference for apartment rentals in terms of time, including desired check-in time and rental duration. Current reservation status indicates whether the apartment is currently booked. If so, detailed information such as the reservation start and end times is included. Historical rental cycles reflect the duration of past rentals.

[0155] In one embodiment, a detailed information table is created for each candidate apartment in the recommendation system's database. This table includes a field for the apartment's marking status (different marking statuses can be represented by specific codes), a time preference field (recording the desired check-in time and rental duration in date format), a current reservation status field (recorded by acquiring and recording reservation information through data interaction with the reservation system), and a historical rental period field (derived from statistics of past rental records). During this step, a database query statement retrieves the corresponding apartment's marking status, time preference, current reservation status, and historical rental period information based on the candidate apartment's unique identifier (e.g., apartment number). This approach leverages the database's structured storage and query capabilities to accurately and efficiently obtain the required information and facilitate data management and maintenance.

[0156] In one embodiment, information acquisition can also be based on a message queue. In the recommendation system, each module communicates with each other through a message queue. When the relevant information of the apartment is updated (such as a change in the marking status, an update in the reservation status, etc.), a message containing the relevant information will be sent to the message queue. The information acquisition module in this step can subscribe to these message queues. When a message about an apartment in the optimized apartment recommendation candidate set is received, the message content is parsed to obtain the apartment's marking status, time preference, current reservation status, and historical rental cycle information. This method is real-time and can obtain the latest information in a timely manner. It is suitable for scenarios where information is frequently updated and requires a timely response.

[0157] S6012: Construct an apartment allocation grid model, map the apartment marking status and time preference into the apartment allocation grid model, and obtain an apartment allocation grid.

[0158] In this step, the apartment allocation grid model is a model structure used to visually represent the allocation of apartment resources. The purpose of building this model is to better analyze the relationship between various apartment states and user needs. Apartment tag status and time preferences are important information closely related to user needs and the current state of the apartment. Mapping these into the apartment allocation grid model allows this information to be integrated and reflected in the model, thus forming the apartment allocation grid.

[0159] In one embodiment, a two-dimensional apartment allocation grid model is constructed, and the details can be referred to the above introduction.

[0160] In one embodiment, a multi-layer apartment allocation grid model can also be constructed. In addition to the time dimension and the apartment attribute dimension, other dimensions are added, such as the user type dimension (for example, office workers, students, etc.) or the demand priority dimension. For each candidate apartment, its apartment marking status and time preference are mapped according to different dimensions. For example, in the user type dimension, if it is an office worker and the apartment is marked as high-intent, it will be represented by a specific identifier in the corresponding layer. Through this multi-layer mapping method, a more complex and comprehensive apartment allocation grid is formed. This method can consider the impact of multiple factors on apartment allocation in more detail, and is suitable for scenarios with more complex requirements for apartment allocation.

[0161] S6013: Calculate the conflict coefficient of the conflicting apartment according to the apartment allocation grid and the apartment reservation status.

[0162] In this step, the apartment allocation grid contains information such as apartment marking status and time preference, and the apartment reservation status reflects the actual booking status of the apartment.

[0163] In one embodiment, a weight can be set for each dimension in the apartment allocation grid and the apartment's reservation status. For example, the popular listings in the apartment tag status are weighted higher, the long-term rental time preference weight is set based on market demand (assuming that long-term rental demand is high, the weight is high), and the apartment reservation status is weighted higher for being booked and about to expire. Then, based on the apartment's position in the grid, the corresponding dimension weights are added together and multiplied by the weight of the apartment's reservation status to obtain the conflict coefficient. For example, Apartment D is in the long-term rental-popular listing area, and the sum of its corresponding dimension weights is 0.6. The apartment reservation status weight is 0.4 (assuming that it is booked but has two days left to expire). The conflict coefficient = 0.6 × 0.4 = 0.24.

[0164] In one embodiment, in the apartment allocation grid, each apartment occupies a certain "area" in the time dimension (for example, an area represented by a combination of time period and other attributes). For conflicting apartments, the overlapping "area" of the apartment in the time dimension with other conflicting apartments is calculated. At the same time, considering the reservation status of the apartment, if the apartment has been booked for part of the time period, this part of the time period is given a higher weight when calculating the overlap (because this part of the time is no longer available, which will increase the degree of conflict). The conflict coefficient is then calculated using a formula, for example, conflict coefficient = overlapping area × (1 + reservation status weight) / total demand area, where the total demand area is the total area of ​​the conflicting apartment involved in the demands of multiple users. This method intuitively quantifies the degree of conflict from the perspective of area overlap and takes into account the impact of the apartment reservation status on the conflict.

[0165] In one embodiment, step S6013 may be implemented as follows:

[0166] Step D: Based on the apartment allocation grid model, locate the time overlapping intervals in the conflicting apartments. The time overlapping intervals are time periods in which multiple users express high interest in the same apartment within the same time period.

[0167] Taking apartment rentals in a certain city as an example, a simple apartment allocation grid model is constructed. Rows represent different urban districts, columns represent different price ranges (e.g., 1,000-2,000 yuan, 2,000-3,000 yuan, etc.), and a time dimension (in weeks) is added. This way, each apartment can be assigned a specific location within this grid based on factors such as its location, price range, and rental period.

[0168] A time overlap interval represents the period of time during which multiple users express high interest in the same apartment. This high interest is reflected in specific user actions on the platform, such as adding an apartment to favorites, frequently viewing apartment details, and checking prices multiple times within a short period of time. When these actions by multiple users occur within a specific time period, this period constitutes a time overlap interval.

[0169] For example, between September 1st and September 5th, four users added Apartment N to their favorites and frequently viewed its details. Therefore, September 1st to September 5th represents an overlapping time period for Apartment N. This period indicates that Apartment N attracted the attention of multiple users during this time period, potentially leading to high competition.

[0170] In one embodiment, apartment-related user interaction data can be extracted from the database of an apartment rental platform, including user operation records and corresponding timestamp information. These operation records primarily indicate high-intent operations, such as favorites and frequently viewed items. A temporary data structure is then created for each apartment to store information about possible time-overlapping intervals. Initially, this data structure is empty. Next, the user operation records are traversed. For each operation record, if the time corresponding to the operation falls within a time interval already stored in the temporary data structure, the relevant statistics for that time interval are updated (e.g., the count of operations within that interval is incremented). If the time corresponding to the operation does not fall within any existing time interval, a new time interval is created, and relevant information about the operation (e.g., operation type, user ID, etc.) is added to this new interval. Finally, all time intervals in the temporary data structure are traversed, and those with an operation count exceeding a preset threshold (e.g., three operations) are identified as time-overlapping intervals for that apartment.

[0171] Step F: Calculate the overlapping time length of the time overlapping interval.

[0172] The overlap duration is a quantitative description of the time period during which multiple users expressed high interest in the same apartment. This duration can be calculated in units of days or hours, depending on your needs.

[0173] For example, if a time overlap period runs from 9:00 AM on October 1st to 5:00 PM on October 3rd, the overlap period is three days when calculated in days. If calculated in hours, converting both the start and end times to hours will yield a more accurate overlap period. A longer overlap period means more users are competing for the apartment during that time period, which has a significant impact on subsequent assessments of apartment conflict and the development of recommendation strategies.

[0174] Step G: Calculating a conflict coefficient based on the overlapping time length, the first weight coefficient of the overlapping time length, the apartment reservation status, and the second weight coefficient of the apartment reservation status.

[0175] For example, a higher conflict coefficient indicates that the apartment faces greater competitive pressure or special leasing circumstances during the recommendation process, and special handling may be required in the recommendation strategy, such as adjusting the recommendation order or providing additional prompts to users.

[0176] The first weighting factor is the weighting factor for the overlap duration. It indicates the relative importance of the overlap duration in calculating the conflict coefficient. It can be a pre-set value, determined based on market analysis, experience, or data analysis. This factor determines the impact of the overlap duration on the conflict coefficient.

[0177] For example, if the first weighting factor is set high, the overlap time will have a greater impact on the conflict assessment of conflicting apartments. For example, suppose there are apartments P and Q with the same reservation status, but apartment P has a longer overlap time period and a higher first weighting factor. Therefore, apartment P will have a higher conflict score than apartment Q.

[0178] The second weighting factor reflects the relative importance of the apartment's reservation status in calculating the conflict coefficient. Similar to the first weighting factor, it can be a pre-set value, determined based on market conditions, experience, or data analysis. This factor determines the degree to which the apartment's reservation status affects the conflict coefficient.

[0179] For example, if the second weight coefficient is large, the apartment's reservation status will have a greater impact on the final result when calculating the conflict coefficient. For example, if Apartment R is booked with one day left for occupancy, and Apartment S is booked with one week left for occupancy, if the second weight coefficient is large, and all other conditions are the same, Apartment R will have a higher conflict coefficient than Apartment S because its reservation status is closer to occupancy and its availability is lower.

[0180] In one embodiment, the conflict coefficient is calculated by using a weighted summation method or the like.

[0181] S602: Based on the conflict coefficient, the optimized apartment recommendation candidate set is screened to obtain a target apartment recommendation candidate set.

[0182] In the embodiment of the present application, the conflict coefficient provides a basis for screening, and whether the apartment is retained in the final recommended candidate set is determined based on the size of the conflict coefficient.

[0183] In one embodiment, a conflict coefficient threshold can be determined based on historical data and experience. For example, a conflict coefficient greater than 0.6 indicates a high level of conflict. Conflicting apartments with conflict coefficients greater than this threshold are removed from the optimized set of recommended apartments. Apartments with conflict coefficients less than or equal to this threshold are retained in the set. This screening process yields the target apartment recommendation set. This approach quickly selects relatively suitable apartments based on the set criteria, reducing competition and conflict between users.

[0184] In one embodiment, reference Figure 8To improve the accuracy of apartment recommendations, step S70 can be implemented by the following steps:

[0185] S701. Perform risk assessment on the recommended apartment in the target apartment recommendation candidate set to obtain an apartment risk index of the recommended apartment. The apartment risk index is determined based on the apartment vacancy rate, tenant default history information, and regional security information of the recommended apartment.

[0186] The apartment vacancy rate refers to the proportion of time an apartment remains unrented within a specific timeframe. For example, if an apartment has been vacant for six months in the past year, its vacancy rate is 50%. A high vacancy rate may indicate issues with the apartment, such as a poor location, excessively high rents, or poor condition, increasing rental risk.

[0187] Tenant default history information refers to whether previous tenants have committed any defaults, such as failing to pay rent on time or terminating their lease early. If an apartment has a high record of tenant defaults, new tenants may also face similar risks when they move in.

[0188] Regional security information refers to the security environment in the area where the apartment is located, including crime rates, security facilities, etc. For example, tenants in areas with poor security may face risks to their property and even their personal safety.

[0189] In one embodiment, to calculate the apartment vacancy rate, the apartment's rental history is collected, the length of vacancy within a specific time period (e.g., the past year) is counted, and then divided by the total length of time to obtain the vacancy rate. For tenant default history information, a tenant credit database can be established. Whether a tenant has committed a default is recorded when they move in and out of the apartment. For each apartment to be recommended, the database is queried for the number of tenant default records associated with the apartment, and a risk score is converted based on a specific algorithm (e.g., the greater the number of default records, the higher the risk score). For regional security information, security data released by local police or relevant departments, such as crime rate data, can be obtained. Based on preset criteria, the security situation is classified into different levels (e.g., high risk, medium risk, low risk) and assigned corresponding risk scores. Finally, a weighted calculation is performed based on pre-set weights (e.g., a weight of 0.3 for the apartment vacancy rate, a weight of 0.3 for tenant default history, and a weight of 0.4 for regional security information) to obtain the apartment risk index for the recommended apartment. This method can comprehensively and objectively assess the risks of the apartments to be recommended, providing an important basis for the subsequent generation of the global optimal recommendation plan, because accurate risk assessment helps avoid recommending high-risk apartments to users and improve users' rental experience and satisfaction.

[0190] In one embodiment, step S701 can be implemented as follows:

[0191] P1. Determine the risk factor value of each apartment in the target apartment recommendation candidate set based on the preset risk assessment table.

[0192] The pre-set risk assessment form is a pre-set, standardized tool for assessing apartment risk. It can include various factors that may affect apartment risk and their corresponding assessment criteria. For example, for the risk factor of apartment vacancy rate, the risk assessment form might specify that if the apartment vacancy rate is between 0-10%, the corresponding risk factor value is 1; if it is between 10%-30%, the risk factor value is 3; if it exceeds 30%, the risk factor value is 5. Regarding tenant default history information, if there is no tenant default record, the risk factor value is 1; if there are 1-2 default records, the risk factor value is 3; if there are more than 2 default records, the risk factor value is 5. Similarly, information on regional public security conditions is similar: areas with good public security have a risk factor value of 1; areas with average public security have a risk factor value of 3; and areas with poor public security have a risk factor value of 5.

[0193] In one embodiment, a database can be created to store the contents of the risk assessment table. This database can exist in the form of a relational database, with each risk factor as a field, and different value ranges corresponding to different risk factor values ​​as records. Then, for each apartment in the target apartment recommendation candidate set, the corresponding apartment vacancy rate, tenant default history information, and regional security information are queried. The queried information is matched with the standards in the risk assessment table to determine the risk factor value of each apartment under each risk factor. This method provides a unified and objective standard for determining the risk factor value of an apartment, making the risk assessment process consistent and comparable. Because different apartments can be evaluated according to the same rules based on the same assessment table, the influence of subjective factors is reduced.

[0194] P2. Quantify the risk factors to obtain the single factor risk value and analyze the correlation between risk factors.

[0195] Quantifying risk factors involves converting previously determined risk factor values ​​into single-factor risk values ​​that can be used for calculations. For example, a risk factor value of 3 might need to be converted to 0.6 (assuming it falls within the 0-1 range) based on a specific quantification algorithm. Analyzing correlations between risk factors is also crucial. For example, there may be a correlation between apartment vacancy rates and tenant default histories. A high apartment vacancy rate might lead landlords to lower their tenant screening standards to rent out the apartment quickly, increasing the likelihood of tenant defaults.

[0196] In one embodiment: For quantitative processing, a linear mapping method can be used. Determine the value range of each risk factor value (such as 1-5) and the target quantitative interval (such as 0-1). Convert the risk factor value into a single factor risk value through a simple linear function (such as y = (x-1) / (1-5), where x is the risk factor value and y is the single factor risk value). For analyzing the correlation between risk factors, a correlation analysis method can be used. Collect a large amount of historical apartment data, including apartment vacancy rates, tenant default history information, and regional security information. Determine the correlation between these data by calculating the correlation coefficient (such as the Pearson correlation coefficient). This method enables risk factors to participate in subsequent calculations in a unified quantitative form, and by analyzing the correlation, it is possible to more comprehensively consider the mutual influence between risk factors, thereby improving the accuracy of risk assessment.

[0197] P3, adjust the risk factor value according to the correlation and single factor risk value to obtain the adjusted risk factor value.

[0198] Because risk factors are correlated, risk factor values ​​need to be adjusted based on this correlation. For example, if a positive correlation is found between apartment vacancy rates and tenant default history information, and the single-factor risk value of apartment vacancy rates is 0.8, and the single-factor risk value of tenant default history information is 0.6, based on pre-set adjustment rules (e.g., if the two are positively correlated and the correlation coefficient is high, a weighted average adjustment is performed on the risk factor values ​​of the two), the risk factor values ​​of apartment vacancy rates might be adjusted to 0.7, and the risk factor value of tenant default history information might be adjusted to 0.7.

[0199] In one embodiment, an adjustment rule is first determined based on the correlations obtained from the previous analysis. If the two risk factors are positively correlated and the correlation coefficient is greater than a certain threshold (e.g., 0.6), a weighted average adjustment method is used. If the two risk factors are negatively correlated and the correlation coefficient is less than a certain threshold (e.g., -0.6), another preset adjustment rule is used (e.g., increasing or decreasing one of the risk factor values ​​by a certain proportion). The risk factor values ​​are then adjusted according to the determined adjustment rule to obtain the adjusted risk factor values.

[0200] The above approach can more accurately reflect the mutual influence between risk factors, avoid viewing each risk factor in isolation, and thus make the risk assessment results more in line with actual conditions.

[0201] P4, calculates the apartment risk index of the recommended apartment in the target apartment recommendation candidate set based on the adjusted risk value and risk factor value.

[0202] The apartment risk index is a comprehensive assessment indicator derived from a comprehensive consideration of all risk factors. The calculation requires a comprehensive consideration of both adjusted and original risk factor values. For example, if the adjusted apartment vacancy rate risk factor is 0.7, the original tenant default history risk factor is 3 (possibly 0.6 after quantitative conversion), and the regional public security risk factor is 1 (0.2 after quantization), then based on pre-set weights (e.g., a weight of 0.3 for apartment vacancy rate, 0.3 for tenant default history, and 0.4 for regional public security), the calculated apartment risk index is 0.7 × 0.3 + 0.6 × 0.3 + 0.2 × 0.4 = 0.47.

[0203] In one embodiment, the weight of each risk factor in calculating the apartment risk index is first determined. These weights can be derived based on experience or data analysis. Then, the adjusted risk value and the risk factor values ​​are weighted and summed according to their respective weights to obtain the apartment risk index. This approach can produce a comprehensive and holistic apartment risk assessment, providing an accurate risk reference for subsequent recommendation decisions and ensuring that the risks of apartments recommended to users are fully considered.

[0204] S702: Generate a global optimal apartment recommendation plan based on the apartment risk index and the user's current demand priority information.

[0205] The apartment risk index reflects the risk profile of the apartment itself, while the user's current priority information reflects the importance they place on various factors, such as rent and location. For example, if a user prioritizes rent compatibility, and a particular apartment has a slightly higher risk index but a rent that is well within their budget, the relationship between the two factors needs to be comprehensively considered to determine whether to include it in the global optimal recommendation.

[0206] In one embodiment, first, the user's current demand priority information is converted into a weight vector of different consideration dimensions (such as rental suitability, geographical convenience, etc.). Then, the various indicators of the recommended apartment (including indicators related to the risk index) are weighted according to these weights. For the risk index, it can be regarded as a special dimension, and the corresponding weight is set according to the user's acceptance of risk (such as risk-averse users will give the risk index a higher weight). The comprehensive score of each apartment to be recommended is calculated, and the apartments with the highest scores constitute the global optimal apartment recommendation plan. This method can meet the user's current needs to the greatest extent while considering the risks of the apartment, and improve the rationality and applicability of the recommendation plan, because the recommendation plan that combines risks and user needs can better meet the user's expectations and reduce the possibility of users rejecting the recommendation.

[0207] In the embodiment of the present application, the global optimal apartment recommendation scheme is an apartment recommendation strategy for users formed after comprehensive consideration of multiple factors. The global optimal apartment recommendation scheme is based on the apartment risk index of the recommended apartment (determined by the apartment vacancy rate, tenant default history information and regional security situation information) and the user's current demand priority information. The apartment risk index reflects the potential risk factors of the apartment itself, covering everything from the apartment's own rental status to the tenant's historical behavior and the security environment of the area. The user's current demand priority information reflects the user's emphasis on different consideration dimensions such as rent adaptability and geographical convenience in a specific period. By quantifying and weighing these factors, and applying specific calculation methods (such as weighted calculations, etc.), we can screen out an apartment combination that achieves the best balance between meeting user needs and controlling risks. This combination is the global optimal apartment recommendation scheme, which aims to provide users with apartment selection suggestions that best meet their needs and have controllable risks.

[0208] S703: Push apartments to users based on the global optimal apartment recommendation plan.

[0209] The optimal global apartment recommendation is the result of careful screening and optimization in the previous steps. It includes the most suitable apartments for the user after considering risk and user needs. For example, if the optimal global recommendation includes three apartments, these three apartments stand out from the other recommended apartments and strike a good balance between meeting user needs and controlling risk.

[0210] In one embodiment, a push queue can be established to place the apartments in the global optimal apartment recommendation plan in descending order according to the comprehensive score (calculated in step S702). Then, according to the push method set by the user (such as email push, mobile phone APP message push, etc.), the apartment information is pushed to the user in sequence according to the queue order. During the push process, some key information, such as the main advantages of the apartment (such as low risk, high cost performance, etc.), can be attached to attract the user's attention. This method can push the most suitable apartment listings to the user in a reasonable and orderly manner, increasing the possibility of the user receiving and selecting the recommended apartment, because the clear and focused push method allows users to quickly understand the advantages of the recommended apartment and make decisions.

[0211] Accordingly, in order to better implement the above method, the embodiment of the present application also provides an apartment rental matching recommendation system based on multi-objective optimization. Figure 9 As shown, the apartment rental matching recommendation system based on multi-objective optimization includes an acquisition module 801, a cluster analysis module 802, a generation module 803, a scoring module 804, an optimization module 805, an identification module 806, and a recommendation module 807, which are specifically as follows:

[0212] Acquisition module 801 is used to obtain user historical behavior data and market demand information, and extract user behavior features from the historical behavior data to obtain key behavior features. The historical behavior data includes at least user browsing history, feedback scores, and interaction time distribution. The market demand information includes at least apartment renewal frequency, supply-demand ratio, and regional popularity index.

[0213] Cluster analysis module 802, configured to perform hierarchical cluster analysis on the market demand information to obtain a market supply and demand grouping model;

[0214] A generating module 803 is configured to generate an initial apartment recommendation candidate set based on the key behavioral characteristics and the market supply and demand grouping model, wherein the initial apartment recommendation candidate set includes at least one candidate apartment;

[0215] Scoring module 804 is configured to calculate a comprehensive score for each candidate apartment in the initial recommended candidate set based on multiple comprehensive consideration dimensions of the candidate apartments, thereby obtaining a comprehensive recommendation score for each candidate apartment in the initial recommended candidate set, wherein the multiple comprehensive consideration dimensions include at least rent compatibility, location convenience, completeness of facilities, and housing condition;

[0216] An optimization module 805 is configured to optimize the initial apartment recommendation candidates based on the comprehensive recommendation score and the user's current demand priority information to obtain an optimized apartment recommendation candidate set;

[0217] an identification module 806 for identifying conflicting apartments in the optimized recommendation candidate set, wherein the conflicting apartments refer to apartments marked as highly preferred by multiple users within the same time period;

[0218] The recommendation module 807 is configured to perform apartment screening on the optimized apartment recommendation candidate set based on the conflicting apartments to obtain a target apartment recommendation candidate set; and perform apartment push processing on the user based on the target apartment recommendation candidate set.

[0219] In one embodiment, the recommendation module 807 is specifically configured to:

[0220] Mapping the conflicting apartments into an apartment allocation grid model, and calculating the conflict coefficient of the conflicting apartments based on the user's time preference and apartment reservation status;

[0221] Based on the conflict coefficient, the optimized apartment recommendation candidate set is screened to obtain a target apartment recommendation candidate set.

[0222] In one embodiment, the recommendation module 807 is specifically configured to:

[0223] Obtaining the apartment marking status and time preference of each candidate apartment in the optimized apartment recommendation candidate set, and querying the current reservation status and historical rental period of the candidate apartment;

[0224] Constructing an apartment allocation grid model, mapping the apartment marking state and time preference to the apartment allocation grid model to obtain an apartment allocation grid;

[0225] Calculating a conflict coefficient of a conflicting apartment according to the apartment allocation grid and the apartment reservation status.

[0226] In one embodiment, the recommendation module 807 is specifically configured to:

[0227] Locating time-overlapping intervals in conflicting apartments based on the apartment allocation grid model, wherein the time-overlapping intervals are time periods in which multiple users express high interest in the same apartment within the same time period;

[0228] Calculating the overlapping time length of the time overlapping interval;

[0229] A conflict coefficient is calculated based on the overlapping time length, a first weight coefficient of the overlapping time length, the apartment reservation status, and a second weight coefficient of the apartment reservation status.

[0230] In one embodiment, the scoring module 804 is configured to: perform weighted processing on the rental suitability, geographical convenience, completeness of facilities, and housing conditions of the candidate apartments to obtain a comprehensive recommendation score for each apartment in the initial recommendation candidate set;

[0231] The optimization module 805 is used to:

[0232] Configuring a dynamic weight adjustment mechanism based on the comprehensive recommendation score;

[0233] Adjust the weights of the comprehensive consideration dimensions of each apartment based on the demand priority information and the dynamic weight adjustment mechanism;

[0234] According to the adjusted weights, the rental suitability, geographical convenience, completeness of facilities and housing conditions of the candidate apartments are weighted to obtain a comprehensive recommendation score for each apartment in the initial recommendation candidate set.

[0235] In one embodiment, the optimization module 805 is configured to:

[0236] Calculating individual scores for rental suitability, geographical convenience, completeness of facilities, and housing condition for each apartment in the initial recommended candidate set;

[0237] Normalizing the individual scores using a nonlinear mapping layer in the multidimensional matching algorithm to obtain standardized individual scores;

[0238] Based on the standardized individual scores, the comprehensive recommendation score is calculated using the weighted fusion layer in the multidimensional matching algorithm; the weighted fusion layer dynamically adjusts the weight of each individual score according to the priority of user needs.

[0239] In one embodiment, the optimization module 805 is configured to:

[0240] Obtaining the user's real-time demand priority information, and parsing the demand priority information to obtain semantic description information of the demand priority information;

[0241] Performing quantitative modeling on the semantic description information to generate a priority weight vector;

[0242] According to the priority weight vector, the weight ratios of the rental suitability, geographical convenience, facility completeness and housing condition are dynamically adjusted.

[0243] In one embodiment, the recommendation module 807 is configured to:

[0244] Performing a risk assessment on the recommended apartment in the target apartment recommendation candidate set to obtain an apartment risk index for the recommended apartment, wherein the apartment risk index is determined based on the apartment vacancy rate, tenant default history information, and regional security information of the recommended apartment;

[0245] Generate the best global apartment recommendation plan based on the apartment risk index and the user's current demand priority information;

[0246] An apartment is pushed to the user according to the global optimal apartment recommendation plan.

[0247] In one embodiment, the recommendation module 807 is configured to:

[0248] Determine the risk factor value of each apartment in the target apartment recommendation candidate set according to a preset risk assessment table;

[0249] Quantifying the risk factors to obtain single-factor risk values, and analyzing the correlations between the risk factors;

[0250] Adjusting the risk factor value according to the correlation and the single factor risk value to obtain an adjusted risk factor value;

[0251] The apartment risk index of the apartment to be recommended in the target apartment recommendation candidate set is calculated according to the adjusted risk value and the risk factor value.

[0252] The implementation of each of the above modules can be specifically referred to the above method embodiments, which will not be described in detail here. The technical effects achieved by each module and device can be referred to the description of the above method embodiments.

[0253] It should be noted that, in specific implementations, the above modules can be arbitrarily combined, integrated into one or more modules, or implemented as independent entities. Furthermore, the above modules can be implemented in the form of hardware or software functional modules. If the integrated modules are implemented in the form of software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. The aforementioned storage medium can be a read-only memory, a magnetic disk, or an optical disk, etc.

[0254] like Figure 10 As shown, an embodiment of the present application further provides a computer device 90, characterized in that it includes a processor 901 and a memory 902, wherein the memory 902 stores a computer program, and when the computer program is executed by the processor 901, the processor 901 performs the steps of any of the methods described above.

[0255] In the several embodiments provided in this application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0256] The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed across multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, the functional units in the various embodiments of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated units may be implemented in the form of hardware or in the form of software functional units.

[0257] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0258] In one aspect, an embodiment of the present application provides a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the method provided in one aspect of the embodiment of the present application.

[0259] The terms "first", "second", etc. in the description, claims and drawings of the embodiments of this application are used to distinguish different objects, rather than to describe a specific order. In addition, the term "including" and any of its variations are intended to cover

[0260] Non-exclusive inclusion. For example, a process, method, apparatus, product, or device comprising a series of steps or modules is not limited to the listed steps or modules, but may optionally include steps or modules not listed, or may optionally include other steps and modules inherent to the process, method, apparatus, product, or device.

[0261] Those skilled in the art will appreciate that the modules and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0262] The above disclosure is only a preferred embodiment of the present application, and certainly cannot be used to limit the scope of rights of the present application. Therefore, equivalent changes made according to the claims of the present application are still within the scope covered by the present application.

Claims

1. A method for apartment rental matching recommendation based on multi-objective optimization, characterized in that: include: Obtaining historical user behavior data and market demand information, extracting user behavior features from the historical behavior data to obtain key behavior features, wherein the historical behavior data includes at least user browsing history, feedback scores, and interaction time distribution, and the market demand information includes at least apartment update frequency, supply-demand ratio, and regional popularity index; Performing hierarchical cluster analysis on the market demand information to obtain a market supply and demand grouping model; generating an initial apartment recommendation candidate set based on the key behavioral characteristics and the market supply and demand grouping model, wherein the initial apartment recommendation candidate set includes at least one candidate apartment; performing a comprehensive score calculation for each candidate apartment in the initial recommended candidate set based on multiple comprehensive consideration dimensions of the candidate apartments, to obtain a comprehensive recommendation score for each candidate apartment in the initial recommended candidate set, wherein the multiple comprehensive consideration dimensions of the candidate apartments include at least rent compatibility, geographical convenience, completeness of facilities, and housing condition; Optimizing the initial apartment recommendation candidates based on the comprehensive recommendation score and the user's current demand priority information to obtain an optimized apartment recommendation candidate set; Identifying conflicting apartments in the optimized recommendation candidate set, wherein the conflicting apartments refer to apartments marked as highly preferred by multiple users within the same time period; Performing apartment screening on the optimized apartment recommendation candidate set based on the conflicting apartments to obtain a target apartment recommendation candidate set; An apartment is pushed to the user according to the target apartment recommendation candidate set.

2. The method according to claim 1, characterized in that The optimized apartment recommendation candidate set is screened based on the conflicting apartments to obtain a target apartment recommendation candidate set, including: Mapping the conflicting apartments into an apartment allocation grid model, and calculating the conflict coefficient of the conflicting apartments based on the user's time preference and apartment reservation status; Based on the conflict coefficient, the optimized apartment recommendation candidate set is screened to obtain a target apartment recommendation candidate set.

3. The method according to claim 2, characterized in that Mapping the conflicting apartments to the apartment allocation grid model and calculating the conflict coefficient of the conflicting apartments based on the user's time preference and apartment reservation status includes: Obtaining the apartment marking status and time preference of each candidate apartment in the optimized apartment recommendation candidate set, and querying the current reservation status and historical rental period of the candidate apartment; Constructing an apartment allocation grid model, mapping the apartment marking state and time preference to the apartment allocation grid model to obtain an apartment allocation grid; Calculating a conflict coefficient of a conflicting apartment according to the apartment allocation grid and the apartment reservation status.

4. The method according to claim 3, characterized in that Calculating the conflict coefficient of the conflicting apartment according to the apartment allocation grid and the apartment reservation status includes: Locating time-overlapping intervals in conflicting apartments based on the apartment allocation grid model, wherein the time-overlapping intervals are time periods in which multiple users express high interest in the same apartment within the same time period; Calculating the overlapping time length of the time overlapping interval; A conflict coefficient is calculated based on the overlapping time length, a first weight coefficient of the overlapping time length, the apartment reservation status, and a second weight coefficient of the apartment reservation status.

5. The method according to any one of claims 1 to 4, characterized in that Based on the multiple comprehensive consideration dimensions of the candidate apartments, a comprehensive score calculation is performed on each candidate apartment in the initial recommendation candidate set to obtain a comprehensive recommendation score for each apartment in the initial recommendation candidate set, including: Weighting the rental suitability, geographical convenience, completeness of facilities, and housing conditions of the candidate apartments to obtain a comprehensive recommendation score for each apartment in the initial recommended candidate set; The initial apartment recommendation candidates are optimized based on the comprehensive recommendation score and the user's current demand priority information to obtain an optimized apartment recommendation candidate set, including: Configuring a dynamic weight adjustment mechanism based on the comprehensive recommendation score; Adjust the weights of the comprehensive consideration dimensions of each apartment based on the demand priority information and the dynamic weight adjustment mechanism; According to the adjusted weights, the rental suitability, geographical convenience, completeness of facilities and housing conditions of the candidate apartments are weighted to obtain a comprehensive recommendation score for each apartment in the initial recommendation candidate set.

6. The method according to claim 5, characterized in that The rental suitability, geographical convenience, completeness of facilities and housing conditions of the candidate apartments are weighted to obtain a comprehensive recommendation score for each apartment in the initial recommendation candidate set, including: Calculating individual scores for rental suitability, geographical convenience, completeness of facilities, and housing condition for each apartment in the initial recommended candidate set; Normalizing the individual scores using a nonlinear mapping layer in a multidimensional matching algorithm to obtain standardized individual scores; Based on the standardized individual scores, the comprehensive recommendation score is calculated using the weighted fusion layer in the multidimensional matching algorithm; the weighted fusion layer dynamically adjusts the weight of each individual score according to the priority of user needs.

7. The method according to claim 5, characterized in that Based on the demand priority information and the dynamic weight adjustment mechanism, the weights of the comprehensive consideration dimensions of each apartment are adjusted, including: Obtaining the user's real-time demand priority information, and parsing the demand priority information to obtain semantic description information of the demand priority information; Performing quantitative modeling on the semantic description information to generate a priority weight vector; According to the priority weight vector, the weight ratios of the rental suitability, geographical convenience, facility completeness and housing condition are dynamically adjusted.

8. The method according to claim 1, characterized in that The performing apartment push processing on the user according to the target apartment recommendation candidate set includes: Performing a risk assessment on the recommended apartment in the target apartment recommendation candidate set to obtain an apartment risk index for the recommended apartment, wherein the apartment risk index is determined based on the apartment vacancy rate, tenant default history information, and regional security information of the recommended apartment; Generate the best global apartment recommendation plan based on the apartment risk index and the user's current demand priority information; An apartment is pushed to the user according to the global optimal apartment recommendation plan.

9. The method according to claim 8, wherein risk assessment is performed on the apartments to be recommended in the target apartment recommendation candidate set to obtain an apartment risk index for the apartments to be recommended, comprising: Determine the risk factor value of each apartment in the target apartment recommendation candidate set according to a preset risk assessment table; Quantifying the risk factors to obtain single-factor risk values, and analyzing the correlations between the risk factors; Adjusting the risk factor value according to the correlation and the single factor risk value to obtain an adjusted risk factor value; The apartment risk index of the apartment to be recommended in the target apartment recommendation candidate set is calculated according to the adjusted risk value and the risk factor value.

10. An apartment rental matching recommendation system based on multi-objective optimization, characterized in that: include An acquisition module is configured to acquire user historical behavior data and market demand information, and extract user behavior features from the historical behavior data to obtain key behavior features. The historical behavior data includes at least user browsing history, feedback scores, and interaction time distribution. The market demand information includes at least apartment renewal frequency, supply-demand ratio, and regional popularity index. A cluster analysis module, configured to perform hierarchical cluster analysis on the market demand information to obtain a market supply and demand grouping model; a generating module, configured to generate an initial candidate set of recommended apartments based on the key behavioral characteristics and the market supply and demand grouping model, wherein the initial candidate set of recommended apartments includes at least one candidate apartment; a scoring module configured to calculate a comprehensive score for each candidate apartment in the initial recommended candidate set based on multiple comprehensive consideration dimensions of the candidate apartments, thereby obtaining a comprehensive recommendation score for each candidate apartment in the initial recommended candidate set, wherein the multiple comprehensive consideration dimensions include at least rent compatibility, geographical convenience, completeness of facilities, and housing condition; an optimization module, configured to optimize the initial apartment recommendation candidates based on the comprehensive recommendation score and the user's current demand priority information to obtain an optimized apartment recommendation candidate set; an identification module, configured to identify conflicting apartments in the optimized recommendation candidate set, wherein the conflicting apartments refer to apartments marked as having high intention by multiple users within the same time period; A recommendation module, configured to perform apartment screening on the optimized apartment recommendation candidate set based on conflicting apartments to obtain a target apartment recommendation candidate set; An apartment is pushed to the user according to the target apartment recommendation candidate set.

Images