House source pushing method based on user portrait

Abstract

The present application relates to the technical field of data processing, and more particularly to a house source pushing method based on user portrait, which constructs a portrait label set by acquiring information labels of the current house of a user; sorts each demand label according to the time interval change of the generation of the user demand label to construct a demand label set; calculates a feature transfer representation coefficient of the demand label set to reorganize the demand label set; determines the demand transfer type of the user by the semantic coincidence of the demand label in the reorganized demand label set and the portrait label in the portrait label set; selects a strategy for preference pushing of the user according to the demand transfer type, and then fuses the continuity demand of the current house experience of the user with the new demand, so as to effectively utilize the evaluation information of the current house of the user to construct the user portrait, accurately anchor the preference features of the user, and improve the matching degree of the pushing result.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method for recommending housing listings based on user profiles. Background Technology

[0002] With the continued acceleration of urbanization in my country, the housing rental and transaction market is expanding. Housing selection is not only related to basic living functions, but also deeply tied to users' living habits and experience preferences. Existing systems generally lack the ability to perceive the dynamic nature of user needs. They fail to effectively utilize users' current housing evaluation information to build user profiles and anchor user preference characteristics, and they also struggle to reflect changes in demand by analyzing user search behavior. Current housing recommendations face pain points such as vague user profiles, lagging demand perception, crude tag processing, and a single push strategy, resulting in low recommendation efficiency and poor user experience. This not only fails to meet users' needs to quickly find suitable housing, but also restricts user retention and transaction conversion on real estate platforms. Against this backdrop, there is an urgent need for a method that can accurately build user profiles, dynamically capture changes in demand, scientifically optimize demand tags, and achieve differentiated pushes, thereby driving housing recommendation systems towards a more accurate and intelligent direction.

[0003] For example, Chinese Patent Publication No. CN113779379A discloses a method, apparatus, device, and storage medium for pushing housing listings based on user profiles. The method includes: acquiring user information of a target user; determining the user type of the target user based on the user information; determining at least one housing tag required by the target user based on the user type; determining first housing information in multiple geographical areas, the first housing information including housing information of at least one house; determining a recommendation score for each house in the first housing information based on the geographical location of each house, at least one housing tag, and the recommendation weight value corresponding to each geographical area; sorting the housing information corresponding to each house in the first housing information according to the recommendation score; and sending the sorted first housing information to the target user's terminal, which can improve the matching degree of housing listings pushed to the target user.

[0004] The following problems still exist in the existing technology:

[0005] Existing technologies do not consider that property listing recommendations need to integrate the continued needs of the current housing experience as well as the new needs arising from various factors. Existing technologies cannot effectively utilize users' current housing evaluation information to build user profiles and cannot anchor user preference characteristics, resulting in homogeneous and low matching results. Summary of the Invention

[0006] To address this, the present invention provides a housing recommendation method based on user profiles, which overcomes the problems of existing technologies being unable to effectively utilize user's current housing evaluation information to construct user profiles, unable to anchor user preference features, resulting in homogeneous recommendation results and low matching degree.

[0007] To achieve the above objectives, the present invention provides a method for recommending housing listings based on user profiles, comprising:

[0008] Obtain information tags about the user's current residence to construct a profile tag set;

[0009] Based on the information tags searched and recorded by users within a preset recording period, several demand tags are determined, and each demand tag is sorted according to the change in the generation time interval of the demand tags to construct a demand tag set.

[0010] The feature transfer representation coefficient of the demand tag set is calculated based on the record information of each demand tag, and the demand tag set is reorganized based on the comparison of the feature transfer representation coefficient with the preset standard.

[0011] Determine the semantic overlap between the demand tags in the reorganized demand tag set and the profile tags in the profile tag set, and determine the user's demand transfer type based on the semantic overlap.

[0012] Based on the demand transfer type, user preferences are pushed to the user. The preference push includes selecting and pushing properties based on profile tags, or determining the explicit properties based on the user's preference for the properties pointed to by the tags, and selecting and pushing properties based on the information tags of the explicit properties.

[0013] Furthermore, the process of constructing a profile tag set includes:

[0014] Retrieve historical evaluation information of the user's current housing, and extract evaluation keywords based on the historical evaluation information;

[0015] The extracted evaluation keywords are identified as information tags, and a set of information tags is identified as a profile tag set based on the order of their frequency of occurrence from most to least.

[0016] Furthermore, the process of constructing a set of demand tags includes:

[0017] Within a preset recording period, search keywords are extracted based on the user's search history, and these search keywords are identified as demand tags.

[0018] The preset recording period is divided into a first recording period and a second recording period, and the generation time of each demand tag is determined in the first recording period and the second recording period, respectively.

[0019] Determine the ratio of the feature quantities generated by each demand tag in the first and second recording periods, and sort the demand tags in descending order of the ratios to construct a demand tag set.

[0020] The generated feature quantity is the average time interval between adjacent generation times of the same demand label in the time dimension.

[0021] Furthermore, the process of calculating the feature transfer representation coefficients of the demand label set includes:

[0022] Extract the frequency of occurrence of each demand tag in the record information and the sorting sequence of each demand tag within the demand tag set;

[0023] Calculate the Pearson correlation coefficient between the frequency of occurrence of each demand tag and the sorted sequence, and determine the calculated Pearson correlation coefficient as the feature transfer characterization coefficient of the demand tag set.

[0024] Furthermore, the process of determining whether to reorganize the set of demand tags includes:

[0025] The feature transfer representation coefficients of the demand tag set are compared with the preset feature transfer representation reference coefficients;

[0026] If the feature transfer characterization coefficient is less than or equal to the feature transfer characterization reference coefficient, then it is determined that the demand label set will not be reorganized.

[0027] If the feature transfer representation coefficient is greater than the feature transfer representation reference coefficient, then it is determined that the demand label set should be reorganized.

[0028] Furthermore, the process of reorganizing the demand tag set includes:

[0029] In response to the determination result of reorganizing the demand tag set, the Pearson correlation coefficient between the occurrence frequency of the demand tags in the first n sort and the first n+1 sort and the sort sequence is calculated respectively.

[0030] If the Pearson correlation coefficient between the frequency of occurrence of the first n+1 sorted demand labels and the sorted sequence is greater than or equal to the Pearson correlation coefficient between the frequency of occurrence of the first n sorted demand labels and the sorted sequence, then the demand label sorted to n is determined as an invalid label.

[0031] The invalid tags are removed from the demand tag set to obtain the reorganized demand tag set.

[0032] Furthermore, the process of determining semantic overlap includes:

[0033] Calculate the semantic similarity between each demand tag in the recombined demand tag set and each portrait tag in the portrait tag set;

[0034] The demand tags and profile tags with semantic similarity greater than a preset similarity threshold are marked as feature non-transfer tag pairs;

[0035] The statistical characteristic is the proportion of the number of non-transfer label pairs to the total number of demand labels.

[0036] Furthermore, the process of pushing user preferences based on the aforementioned demand transfer type includes:

[0037] If the ratio value is greater than the preset ratio reference value, the user's demand transfer type is determined to be the first demand transfer type, and the preference push is to select housing resources based on profile tags for push.

[0038] If the ratio value is less than or equal to the preset ratio reference value, the user's demand transfer type is determined to be the second demand transfer type. The preference push is to determine the tagged explicit properties based on the user's preference for the properties pointed to by the tag, and select properties for push based on the information tags of the tagged explicit properties.

[0039] Furthermore, the process of selecting and pushing properties based on profile tags includes:

[0040] Obtain information tags for candidate properties within a geographic distance range based on the user's current location;

[0041] Candidate properties are selected where the information tag content is completely consistent with the profile tag content, and the sorting sequence of each information tag is completely consistent with the sorting sequence of the profile tag.

[0042] The selected candidate properties will be pushed to users.

[0043] Furthermore, the process of selecting and pushing properties based on the information tags of the explicitly displayed properties includes:

[0044] The cumulative viewing time of users for properties linked to each tag is recorded as the preference score;

[0045] The properties with the highest preference ratings are identified as those with explicit tags.

[0046] Define a geographic distance range based on the user's current location, and retrieve candidate properties within the geographic distance range;

[0047] Extract the information tags and sorting sequence of each information tag from the candidate properties, and filter out the candidate properties whose information tag content is completely consistent with the information tag content of the properties with explicit tags, and whose sorting sequence of each information tag is completely consistent with the information tag sorting sequence of the properties with explicit tags.

[0048] The selected candidate properties will be pushed to users.

[0049] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention constructs a profile tag set by acquiring information tags of the user's current housing; sorts each demand tag according to the generation time interval of the user's demand tags to construct a demand tag set; calculates the feature transfer representation coefficient of the demand tag set to reorganize the demand tag set; determines the user's demand transfer type by the semantic overlap between the demand tags in the reorganized demand tag set and the profile tags in the profile tag set; selects a strategy for user preference push based on the demand transfer type; and further, integrates the continuous needs and new needs of the user's current housing experience, thereby effectively utilizing the user's current housing evaluation information to construct a user profile, accurately anchoring user preference features, and improving the matching degree of push results.

[0050] Furthermore, this invention extracts search keywords from user search records within a preset time period as demand tags. These demand tags are the direct carriers of users' immediate needs. By dividing the preset recording period into two equal recording periods, the smaller the average time interval between adjacent generation times, the more frequently the tag is mentioned. By comparing the ratio of the average time interval between adjacent generation times of the same tag in the two periods, the strengthening or weakening of demand can be accurately perceived. The demand tag set is constructed by sorting the tags from largest to smallest according to this ratio, so that the tags at the top of the set reflect the user's most core and fastest-growing needs.

[0051] Furthermore, this invention reflects the degree to which users have long-term attention to the needs corresponding to the tags by extracting the cumulative frequency of each demand tag in the search records; it reflects the demand ranking of the demand tag by extracting the ranking sequence of the demand tags within the demand tag set; and it obtains the feature transfer characterization coefficient by calculating the Pearson correlation coefficient of these two indicators. The magnitude of this coefficient directly reflects the degree of correlation between the long-term attention to the demand and the demand ranking. If the feature transfer characterization coefficient is within a preset standard range, it indicates that the tag ranking matches the actual importance of the demand, and the demand tag set is effective. If the coefficient exceeds a preset threshold, it indicates that there is a deviation in the tag ranking, and the ranking of some tags does not match their actual attention.

[0052] Furthermore, this invention calculates the semantic similarity between each tag in the recombined demand tag set and each tag in the profile tag set using a natural language processing model. Tag pairs with similarity exceeding a preset threshold are marked as non-transferable feature tag pairs. Then, the proportion of non-transferable feature tag pairs to the total number of demand tags is calculated. This allows for the differentiation of user demand transfer types. When the proportion is greater than a reference value, it indicates that the user's demand is highly consistent with their current housing preferences, and the push strategy anchors the profile tags to match housing resources. If the proportion is less than or equal to the reference value, it indicates that the user's demand has deviated from their current housing preferences, and the push strategy shifts to filtering explicit housing resources based on the user's preference for the housing resources indicated by the tags, and then matching the corresponding housing resources. This achieves precise adaptation between the push strategy and user demand.

[0053] Furthermore, this invention achieves highly relevant property recommendations by anchoring to users' verified residential preferences, adapting to scenarios where user needs have not shifted but rather shifted from the primary need type. It defines a geographical distance range based on the user's current location and retrieves candidate properties, aligning with the user's implicit needs for current living amenities. Secondly, by extracting information tags and ranking sequences from candidate properties and performing a dual comparison of content and ranking with user profile tags, complete consistency in information tag content ensures that candidate properties possess residential attributes long-term recognized by the user, while complete consistency in ranking sequences guarantees that the priority of these attributes highly matches the user's existing residential preferences. Finally, candidate properties that meet both matching criteria are pushed to the user, improving the matching accuracy of the recommendations.

[0054] Furthermore, this invention records the cumulative viewing time of users for properties linked to each tag as a preference score. The principle is that users pay significantly more attention to properties they are interested in than to ordinary properties. The quantification of cumulative viewing time can effectively avoid interference from invalid behaviors such as accidental clicks on preference determination. Then, the property linked to the tag with the highest preference score is directly identified as the tag-explicit property. This property serves as a concrete carrier of the user's most core housing needs at the current stage, and its information tags and the sorting sequence of each information tag become the benchmark for subsequent candidate property selection. Finally, the information tags and sorting sequence of candidate properties are extracted, and a dual screening is performed from two dimensions: tag content and sorting sequence. Completely consistent tag content ensures that candidate properties possess all the core attributes that users care about, and completely consistent sorting sequences ensure that the priority of these attributes is highly consistent with user preferences. Finally, the highly matched properties selected are pushed to users, improving the matching degree of the push results. Attached Figure Description

[0055] Figure 1 This diagram illustrates the execution steps of the user profile-based housing listing recommendation method according to an embodiment of the present invention.

[0056] Figure 2 A flowchart illustrating the steps for constructing a set of demand tags in an embodiment of the present invention;

[0057] Figure 3 This is a flowchart illustrating the logic of determining whether to reorganize the set of demand tags in an embodiment of the present invention.

[0058] Figure 4 This is a diagram illustrating the steps of reorganizing the demand tag set according to an embodiment of the present invention;

[0059] Figure 5 This is a flowchart illustrating the steps for determining semantic overlap in an embodiment of the present invention.

[0060] Figure 6 This is a flowchart illustrating the steps of selecting and pushing properties based on the information tags of the properties displayed in the label, according to an embodiment of the present invention. Detailed Implementation

[0061] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0062] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0063] It should be noted that in the description of this invention, the terms "upper," "lower," "inner," "outer," etc., which indicate the direction or positional relationship, are based on the direction or positional relationship shown in the drawings. This is only for the convenience of description and is not intended to indicate or imply that the device or element must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation of this invention.

[0064] Please see Figure 1 The diagram illustrates the execution steps of the user profile-based property listing push method according to an embodiment of the present invention. The user profile-based property listing push method of the present invention includes:

[0065] Step S100: Obtain information tags of the user's current residence to construct a profile tag set;

[0066] This invention does not limit the method of obtaining the information tag of the user's current housing. It retrieves the historical evaluation text of the user's current housing, uses the jieba word segmentation algorithm to segment the historical evaluation text, and determines the obtained word segmentation as the information tag of the user's current housing.

[0067] Step S200: Determine several demand tags based on the information tags searched and recorded by the user within a preset recording period, and sort each demand tag according to the change in the generation time interval of the demand tags to construct a demand tag set.

[0068] In this invention, the preset recording period can be set by a technician. To avoid insufficient information in the search records due to an excessively short preset recording period, and to avoid redundant information in the search records due to an excessively long preset recording period, preferably, the preset recording period is 60 days.

[0069] Step S300: Calculate the feature transfer representation coefficient of the demand tag set based on the record information of each demand tag, and reorganize the demand tag set based on the comparison of the feature transfer representation coefficient with the preset standard;

[0070] Step S400: Determine the semantic overlap between the demand tags in the reorganized demand tag set and the portrait tags in the portrait tag set, and determine the user's demand transfer type based on the semantic overlap.

[0071] Step S500: Push preferences to users according to the demand transfer type. The preference push includes selecting and pushing properties based on profile tags, or determining the tagged explicit properties based on the user's preference for the tagged explicit properties, and selecting and pushing properties based on the information tags of the tagged explicit properties.

[0072] Among them, the first sorting requirement tags corresponding to each property are different.

[0073] This invention also provides a computer-readable storage medium storing computer program code. When the computer program code is run on a computer, the computer executes the aforementioned method steps to implement the aforementioned method for pushing housing listings based on user profiles.

[0074] Specifically, the process of constructing a set of profile tags includes:

[0075] Retrieve historical evaluation information of the user's current housing, and extract evaluation keywords based on the historical evaluation information;

[0076] The extracted evaluation keywords are identified as information tags, and a set of information tags is identified as a profile tag set based on the order of their frequency of occurrence from most to least.

[0077] In this invention, to ensure that the extracted information tags are representative, the number of times each information tag appears repeatedly in the profile tag set is greater than or equal to 5% of the total number of evaluation keywords in the historical evaluation information.

[0078] The historical evaluation information in this invention can include information such as descriptions of residents' living experience.

[0079] In this invention, if the number of times the information tags appear is the same, the information tags are sorted according to the time order of their first appearance.

[0080] Understandably, users' historical evaluations of their existing housing are genuine preference feedback validated through long-term residence. By retrieving this evaluation information and extracting key evaluation terms, and then visualizing them as information tags, the frequency of these tags in users' historical evaluations of their current housing directly reflects their level of attention and preference for the corresponding residential attributes. For example, the more frequently "near the subway" and "north-south facing" appear, the higher the priority of these attributes in the user's living experience. By sorting the extracted information tags from most frequent to least frequent, a hierarchical arrangement from core preferences to secondary preferences is formed, ensuring that the profile tag set clearly presents the primary and secondary relationships of user preferences. Constructing the profile tag set based on tag frequency ensures a strong correlation between the tags and the user's actual needs.

[0081] Please see Figure 2 The diagram illustrates the steps involved in constructing a demand tag set according to an embodiment of the present invention. The process of constructing the demand tag set includes:

[0082] Step S201: Extract search keywords based on the user's search history within a preset recording period, and determine the search keywords as demand tags;

[0083] In this invention, to ensure that the extracted search keywords are representative, the number of times the search keywords corresponding to the demand tags appear repeatedly is greater than or equal to 5% of the total number of search keywords.

[0084] Step S202: Divide the preset recording period into a first recording period and a second recording period, and determine the generation time of each demand tag in the first recording period and the second recording period respectively;

[0085] For example, the preset 60-day recording period is divided into a first recording period of 30 days and a second recording period of 30 days.

[0086] Step S203: Determine the ratio of the feature quantities generated by each demand tag in the first recording period and the second recording period respectively, and sort the demand tags in descending order of the ratio to construct a demand tag set;

[0087] The generated feature quantity is the average time interval between adjacent generation times of the same demand label in the time dimension.

[0088] In this invention, the demand tags are sorted in descending order of their ratios. If the ratios of multiple demand tags are equal, they are sorted in descending order of their total frequency of occurrence within a preset recording period. If the total frequency of occurrence is still equal, they are sorted in ascending order of the average time interval between occurrences of each demand tag within a second recording period, thus constructing a unique and ordered set of demand tags.

[0089] This invention extracts search keywords from user search records within a preset time period as demand tags. These demand tags directly represent users' immediate needs. By dividing the preset recording period into two equal recording periods, and locating the generation time of each demand tag within the two periods, the average time interval between adjacent generation time points of the tag within each period is calculated. The smaller the average time interval between adjacent generation time points, the higher the frequency of the tag being mentioned. By comparing the ratio of the average time interval between adjacent generation time points of the same tag in the two periods, the strengthening or weakening of demand can be accurately perceived. That is, the larger the ratio, the more significant the increase in the mention frequency of the tag from the early to the later period, and the more obvious the demand. The demand tags are sorted from largest to smallest according to this ratio to construct a set, so that the tags at the top of the set reflect the user's most core and fastest growing needs.

[0090] Specifically, the process of calculating the feature transfer representation coefficients of the demand label set includes:

[0091] Extract the frequency of occurrence of each demand tag in the record information and the sorting sequence of each demand tag within the demand tag set;

[0092] Calculate the Pearson correlation coefficient between the frequency of occurrence of each demand tag and the sorted sequence, and determine the calculated Pearson correlation coefficient as the feature transfer characterization coefficient of the demand tag set.

[0093] Those skilled in the art should understand that the sorting sequence of demand tags in this invention is constructed based on the ratio of the generated feature quantities in two time periods, reflecting the trend of demand becoming stronger or weaker. The frequency of tag occurrence reflects the long-term attention to demand. Both revolve around the dimension of demand importance and have a clear potential correlation logic. Tags with stronger demand are often accompanied by higher long-term attention, while weaker demand leads to reduced attention. This correlation shows a significant linear trend characteristic, so the Pearson correlation coefficient can be used to characterize the correlation strength.

[0094] For example, a set of demand tags containing 6 demand tags is constructed based on user A. The frequency of occurrence and sorting sequence of each tag are shown in the table below:

[0095]

[0096] The feature transfer coefficient r was calculated using the Pearson correlation coefficient formula: r = (6 × 220 - 87 × 21) / [(6 × 1697 - 87] 2 )×(6×91-21 2 )] 1 / 2 =-0.968;

[0097] The feature transfer coefficient r is -0.968, indicating that the frequency of occurrence of each demand tag is negatively correlated with the ranking sequence. This invention does not limit the Pearson correlation coefficient formula, which is an existing mathematical formula. The range of the Pearson correlation coefficient is [-1, 1]. The closer the value is to 1, the more obvious the positive correlation between the two. The closer the value is to -1, the more obvious the negative correlation between the two. The closer the value is to 0, the more obvious the correlation between the two. The Pearson correlation coefficient is a commonly used method in mathematical calculation, which will not be elaborated here.

[0098] Please see Figure 3 As shown, this is a flowchart illustrating the logic of determining whether to reorganize the demand tag set according to an embodiment of the present invention. The process of determining whether to reorganize the demand tag set includes:

[0099] The feature transfer representation coefficients of the demand tag set are compared with the preset feature transfer representation reference coefficients;

[0100] If the feature transfer characterization coefficient is less than or equal to the feature transfer characterization reference coefficient, then it is determined that the demand label set will not be reorganized.

[0101] If the feature transfer representation coefficient is greater than the feature transfer representation reference coefficient, then it is determined that the demand label set should be reorganized.

[0102] In this invention, the preset feature transfer representation reference coefficient can be calculated in advance based on tests. Several user demand tag sets are pre-tested and recorded. The average value of the feature transfer representation reference coefficients corresponding to all user demand tag sets is calculated based on the frequency of occurrence of demand tags in each user's demand tag set and the sorting sequence data. The average value of the calculated feature transfer representation reference coefficients is determined as the feature transfer representation reference coefficient. In this invention, the value range of the feature transfer representation reference coefficient is [-0.93, -0.9]. Preferably, the value of the feature transfer representation reference coefficient is -0.92.

[0103] Understandably, this invention reflects the degree to which users have long-term attention to the needs corresponding to the tags by extracting the cumulative frequency of each demand tag in the search records; it reflects the demand ranking of the demand tag by extracting the ranking sequence of the demand tags within the demand tag set; and it obtains the feature transfer characterization coefficient by calculating the Pearson correlation coefficient of these two indicators. The magnitude of this coefficient directly reflects the degree of correlation between the long-term attention to the demand and the demand ranking. If the feature transfer characterization coefficient is within a preset standard range, it indicates that the tag ranking matches the actual importance of the demand, and the demand tag set is effective. If the coefficient exceeds a preset threshold, it indicates that there is a deviation in the tag ranking, and the ranking of some tags does not match their actual attention.

[0104] Please see Figure 4 The diagram illustrates the steps of reorganizing the demand tag set according to an embodiment of the present invention. The process of reorganizing the demand tag set includes:

[0105] Step S301: In response to the determination result of reorganizing the demand tag set, calculate the Pearson correlation coefficient between the occurrence frequency of the demand tags in the first n sort and the first n+1 sort and the sort sequence.

[0106] Step S302: If the Pearson correlation coefficient between the frequency of occurrence of the first n+1 sorted demand labels and the sorted sequence is greater than or equal to the Pearson correlation coefficient between the frequency of occurrence of the first n sorted demand labels and the sorted sequence, then the demand label sorted as n is determined as an invalid label.

[0107] Step S303: Delete the invalid tags from the demand tag set to obtain the reorganized demand tag set.

[0108] In this invention, to ensure the validity of the Pearson correlation coefficient calculation, n is set to a positive integer greater than 3.

[0109] For example, a set of requirement tags containing 6 requirement tags is constructed for user B. The frequency of occurrence and sorting sequence of each tag are shown in the table below:

[0110]

[0111] The feature transfer representation reference coefficient for the demand tag set corresponding to user B is calculated to be -0.895, which is greater than the feature transfer representation reference coefficient of -0.92. Therefore, it is determined that the demand tag set of user B should be reorganized.

[0112] Compare the frequency of occurrence of the first n sorted and the first n+1 sorted demand tags in the demand tag set corresponding to user B with the Pearson correlation coefficient of the sorted sequence. Since n is a positive integer greater than 3, when n=4, calculate the Pearson correlation coefficient of the first 4 and the first 5 demand tags. The Pearson correlation coefficient of the first 4 demand tags is -0.940, and the Pearson correlation coefficient of the first 5 demand tags is -0.967. Since -0.967 < -0.940, the first 5 demand tags are valid demand tags. Continue to calculate the Pearson correlation coefficient of the first 6 demand tags, which is -0.895. Since -0.895 > -0.940, the tag with the sorted sequence of 5 is determined to be an invalid tag and is deleted. The remaining demand tags are reorganized to obtain the reorganized demand tag set.

[0113] Understandably, after triggering the reorganization decision, the first n and the first n+1 sorted demand tags are selected progressively. The Pearson correlation coefficient between the frequency of occurrence of the two sets of tags and the sorted sequence is calculated. If the Pearson correlation coefficient increases after adding the (n+1)th tag, it means that the nth tag has disrupted the correlation of the original tag set and is an invalid tag. Then, the invalid tag is removed from the set. Through this method of successively verifying and removing interference items, an optimized reorganized demand tag set is obtained, ensuring that the tags in the set can accurately focus on the user's real core needs.

[0114] Please see Figure 5 The diagram illustrates the steps for determining semantic overlap in an embodiment of the present invention. The process for determining semantic overlap includes:

[0115] Step S401: Calculate the semantic similarity between each demand tag in the recombined demand tag set and each portrait tag in the portrait tag set;

[0116] Step 402: Mark the demand tags and portrait tags with semantic similarity greater than a preset similarity threshold as feature non-transfer tag pairs;

[0117] Step S403: Calculate the proportion of the number of non-transfer label pairs to the total number of required labels.

[0118] This invention does not limit the calculation method of semantic similarity. The normalized result of Euclidean distance can be used to represent the similarity. The normalized result of Euclidean distance is in the range of [0, 1]. The closer the value is to 1, the higher the semantic overlap. The relevant algorithms for semantic similarity calculation are commonly used in semantic recognition and word segmentation statistics, and will not be elaborated here.

[0119] In this invention, the preset similarity threshold is set by technicians according to the requirements of screening accuracy. The range of the similarity threshold is [0.88, 0.92]. Preferably, the similarity threshold is 0.9.

[0120] Specifically, the process of pushing user preferences based on the aforementioned demand shift type includes:

[0121] If the ratio value is greater than the preset ratio reference value, the user's demand transfer type is determined to be the first demand transfer type, and the preference push is to select housing resources based on profile tags for push.

[0122] If the ratio value is less than or equal to the preset ratio reference value, the user's demand transfer type is determined to be the second demand transfer type. The preference push is to determine the tagged explicit properties based on the user's preference for the properties pointed to by the tag, and select properties for push based on the information tags of the tagged explicit properties.

[0123] In this invention, the purpose of the preset ratio reference value is to distinguish the relationship between the user's current needs and current experience. Based on this purpose, the range of the ratio reference value is set to [0.65, 0.75], and preferably, the ratio reference value is 0.7.

[0124] Understandably, this invention calculates the semantic similarity between each tag in the recombined demand tag set and each tag in the profile tag set using a natural language processing model. Tag pairs with similarity exceeding a preset threshold are marked as non-transferable feature tag pairs. Then, the proportion of non-transferable feature tag pairs to the total number of demand tags is calculated. This allows for the differentiation of user demand transfer types. When the proportion is greater than a reference value, it indicates that the user's demand is highly consistent with their current housing preferences, and the push strategy anchors the profile tags to match housing resources. If the proportion is less than or equal to the reference value, it indicates that the user's demand has deviated from their current housing preferences, and the push strategy shifts to filtering explicit housing resources based on the user's preference for the housing resources indicated by the tags, and then matching the corresponding housing resources. This achieves precise adaptation between the push strategy and user demand.

[0125] Specifically, the process of selecting and pushing properties based on profile tags includes:

[0126] Obtain information tags for candidate properties within a geographic distance range based on the user's current location;

[0127] Candidate properties are selected where the information tag content is completely consistent with the profile tag content, and the sorting sequence of each information tag is completely consistent with the sorting sequence of the profile tag.

[0128] The selected candidate properties will be pushed to users.

[0129] In this invention, the geographical distance range is set to 5km, that is, the geographical distance range is defined as the area with the user's current location as the center and a radius of 5km.

[0130] Understandably, this invention achieves highly relevant property recommendations by anchoring to users' verified residential preferences. It adapts to scenarios where users' needs have not shifted, specifically the first type of demand shift. It defines a geographical distance range based on the user's current location and retrieves candidate properties, aligning with the user's implicit needs for current living amenities. Secondly, by extracting information tags and ranking sequences from candidate properties and comparing them with profile tags in both content and ranking, complete consistency in information tag content ensures that candidate properties possess residential attributes that users have long recognized. Complete consistency in ranking sequences ensures that the priority of these attributes highly matches the user's existing residential preferences. Finally, candidate properties that meet both matching criteria are pushed to the user, improving the matching accuracy of the recommendation results.

[0131] Please see Figure 6 The diagram illustrates the steps of selecting and pushing properties based on the information tags of explicitly listed properties according to an embodiment of the present invention. The process of selecting and pushing properties based on the information tags of explicitly listed properties includes:

[0132] Step S501: Record the cumulative viewing time of the user on each property listed under each tag as the preference level;

[0133] Step S502: The property with the highest preference level is identified as the property with the most explicit label.

[0134] Step S503: Define a geographical distance range based on the user's current location, and retrieve candidate properties within the geographical distance range;

[0135] Step S504: Extract the information tags of the candidate properties and the sorting sequence of each information tag, and filter out the candidate properties whose information tag content is completely consistent with the information tag content of the property with explicit tags, and whose sorting sequence of each information tag is completely consistent with the information tag sorting sequence of the property with explicit tags.

[0136] Step S505: Push the selected candidate properties to the user.

[0137] In this invention, the first sorting requirement tags corresponding to each property are different. That is, the contents of the first sorting sequence in the set of requirement tags for the properties are "within 500 meters of the subway station", "south-facing living room", "fully furnished", "educational resources" and "top floor with attic". Based on this, five properties are first pushed, and the properties with explicit tags are determined according to the user's preference for these five properties.

[0138] Understandably, recording the cumulative viewing time of users on properties linked to by each tag as a preference score is based on the principle that users will spend significantly more time focusing on properties they are interested in than on ordinary properties. Quantifying the cumulative viewing time effectively avoids interference from invalid behaviors such as accidental clicks on preference determination. Next, the property linked to by the tag with the highest preference score is directly identified as the tagged explicit property. This property serves as a concrete representation of the user's core housing needs at the current stage, and its information tags and their ranking sequence become the benchmark for subsequent candidate property selection. Finally, the information tags and ranking sequences of candidate properties are extracted, and a dual screening is performed from both tag content and ranking sequence dimensions. Completely consistent tag content ensures that candidate properties possess all the core attributes that users care about, while completely consistent ranking sequences ensure that the priority of these attributes highly matches user preferences. Ultimately, the highly matched properties selected are pushed to users, improving the matching accuracy of the push results.

[0139] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

[0140] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for recommending housing listings based on user profiles, characterized in that, include: Obtain information tags about the user's current residence to construct a profile tag set; Based on the information tags searched and recorded by users within a preset recording period, several demand tags are determined, and each demand tag is sorted according to the change in the generation time interval of the demand tags to construct a demand tag set. The feature transfer representation coefficient of the demand tag set is calculated based on the record information of each demand tag, and the demand tag set is reorganized based on the comparison of the feature transfer representation coefficient with the preset standard. The process of calculating the feature transfer characterization coefficient of the demand tag set includes extracting the frequency of occurrence of each demand tag in the record information and the sorting sequence of each demand tag in the demand tag set; Calculate the Pearson correlation coefficient between the frequency of occurrence of each demand tag and the sorting sequence, and determine the calculated Pearson correlation coefficient as the feature transfer characterization coefficient of the demand tag set; The process of determining whether to reorganize the demand tag set includes comparing the feature transfer representation coefficients of the demand tag set with the preset feature transfer representation reference coefficients; If the feature transfer characterization coefficient is less than or equal to the feature transfer characterization reference coefficient, then it is determined that the demand label set will not be reorganized. If the feature transfer representation coefficient is greater than the feature transfer representation reference coefficient, then it is determined that the demand label set should be reorganized. Determine the semantic overlap between the demand tags in the reorganized demand tag set and the profile tags in the profile tag set, and determine the user's demand transfer type based on the semantic overlap. Based on the demand transfer type, user preferences are pushed to the user. The preference push includes selecting and pushing properties based on profile tags, or determining the explicit properties based on the user's preference for the properties pointed to by the tags, and selecting and pushing properties based on the information tags of the explicit properties.

2. The method for recommending housing listings based on user profiles according to claim 1, characterized in that, The process of constructing a profile tag set includes: Retrieve historical evaluation information of the user's current housing, and extract evaluation keywords based on the historical evaluation information; The extracted evaluation keywords are identified as information tags, and a set of information tags is identified as a profile tag set based on the order of their frequency of occurrence from most to least.

3. The method for recommending housing listings based on user profiles according to claim 2, characterized in that, The process of constructing a set of requirement tags includes: Within a preset recording period, search keywords are extracted based on the user's search history, and these search keywords are identified as demand tags. The preset recording period is divided into a first recording period and a second recording period, and the generation time of each demand tag is determined in the first recording period and the second recording period, respectively. Determine the ratio of the feature quantities generated by each demand tag in the first and second recording periods, and sort the demand tags in descending order of the ratios to construct a demand tag set. The generated feature quantity is the average time interval between adjacent generation times of the same demand label in the time dimension.

4. The method for recommending housing listings based on user profiles according to claim 1, characterized in that, The process of reorganizing the demand tag set includes: In response to the determination result of reorganizing the demand tag set, the Pearson correlation coefficient between the occurrence frequency of the demand tags in the first n sort and the first n+1 sort and the sort sequence is calculated respectively. If the Pearson correlation coefficient between the frequency of occurrence of the first n+1 sorted demand labels and the sorted sequence is greater than or equal to the Pearson correlation coefficient between the frequency of occurrence of the first n sorted demand labels and the sorted sequence, then the demand label sorted to n is determined as an invalid label. The invalid tags are removed from the demand tag set to obtain the reorganized demand tag set.

5. The method for recommending housing listings based on user profiles according to claim 4, characterized in that, The process of determining semantic overlap includes: Calculate the semantic similarity between each demand tag in the recombined demand tag set and each portrait tag in the portrait tag set; The demand tags and profile tags with semantic similarity greater than a preset similarity threshold are marked as feature non-transfer tag pairs; The statistical characteristic is the proportion of the number of non-transfer label pairs to the total number of demand labels.

6. The method for recommending housing listings based on user profiles according to claim 5, characterized in that, The process of pushing user preferences based on the aforementioned demand shift type includes: If the ratio value is greater than the preset ratio reference value, the user's demand transfer type is determined to be the first demand transfer type, and the preference push is to select housing resources based on profile tags for push. If the ratio value is less than or equal to the preset ratio reference value, the user's demand transfer type is determined to be the second demand transfer type. The preference push is to determine the tagged explicit properties based on the user's preference for the properties pointed to by the tag, and select properties for push based on the information tags of the tagged explicit properties.

7. The method for recommending housing listings based on user profiles according to claim 6, characterized in that, The process of selecting and pushing properties based on profile tags includes: Obtain information tags for candidate properties within a geographic distance range based on the user's current location; Candidate properties are selected based on the fact that the information tag content is completely consistent with the profile tag content, and the sorting sequence of each information tag is completely consistent with the sorting sequence of the profile tag. The selected candidate properties will be pushed to users.

8. The method for recommending housing listings based on user profiles according to claim 6, characterized in that, The process of selecting and pushing properties based on the information tags of the explicitly displayed properties includes: The cumulative viewing time of users for properties linked to each tag is recorded as the preference score; The properties with the highest preference ratings are identified as those with explicit tags. Define a geographic distance range based on the user's current location, and retrieve candidate properties within the geographic distance range; Extract the information tags and sorting sequence of each information tag from the candidate properties, and filter out the candidate properties whose information tag content is completely consistent with the information tag content of the properties with explicit tags, and whose sorting sequence of each information tag is completely consistent with the information tag sorting sequence of the properties with explicit tags. The selected candidate properties will be pushed to users.

Images