A housing source push method, device, and equipment

Abstract

The present invention discloses a method, device, and apparatus for pushing listings. The method comprises: obtaining a user's click behavior data, search behavior data, chat data, and timeliness data in a house-hunting app; extracting the user's complete preference information from the click behavior data, search behavior data, and chat data; calculating a weight score for each preference information based on the operation weight and timeliness weight of each of the click behavior data, search behavior data, and chat data; extracting target preference information from the complete preference information based on the weight score, and pushing target listings to the user based on the target preference information. The technical solution provided by the present invention improves the accuracy of listing push notifications.

Description

Technical Field

[0001] The present invention relates to the field of software message push, and in particular to a housing listing push method, device and equipment. Background Art

[0002] With the increasing demand for renting and buying properties, housing search software, to meet the personalized needs of different users, typically analyzes users' preferences based on their clicks, searches, and chats within the app, and then recommends suitable properties. However, existing analysis methods lack accuracy, resulting in inaccurate recommended properties. Improving the accuracy of recommended properties is an urgent issue. Summary of the Invention

[0003] In view of this, embodiments of the present invention provide a method, apparatus, and device for pushing house listings, thereby improving the accuracy of house listing push.

[0004] According to a first aspect, the present invention provides a method for pushing housing resources, the method comprising: obtaining click behavior data, search behavior data, chat data and timeliness data of a user in a housing-hunting software, the timeliness data being used to count the time length from the click behavior data, the search behavior data and the chat data to the current moment; extracting all preference information of the user from the click behavior data, the search behavior data and the chat data; calculating a weight score of each preference information based on the operation weight and timeliness weight of each of the click behavior data, the search behavior data and the chat data, the operation weight being the credibility of each of the click behavior data, the search behavior data and the chat data in representing the user's preference, the timeliness weight being generated based on the timeliness data and being used to represent the amount of change in the operation weight over time; extracting target preference information from all preference information based on the weight score, and pushing target housing resources to the user based on the target preference information.

[0005] Optionally, the acquisition of the user's click behavior data, search behavior data, chat data and timeliness data includes: acquiring the user's historical records of browsing, sharing, collecting, booking and signing preview properties in the house-hunting software, and generating the click behavior data based on the preset intention tags marked on the properties in the historical records; converting the user's search terms in the house-hunting software into preset intention tags, and generating the search behavior data based on the converted preset intention tags; extracting the intent keywords and the emotional orientation of the intent keywords in the chat records of the user in the house-hunting software, and converting the intent keywords into preset intention tags with emotional orientation, and then generating the chat data based on the converted preset intention tags, the emotional orientation is used to determine whether the intent keywords are the user's true preference; acquiring the time length from the moment when the click behavior data, the search behavior data and the chat data each behavior occurs to the current moment, and generating the timeliness data based on the time length.

[0006] Optionally, the calculation of the weight score of each preference information based on the respective operation weights and time-limit weights of the click behavior data, search behavior data, and chat data includes: obtaining a current preset intention label corresponding to the current preference information; obtaining a first operation weight and a first time-limit weight of the current preset intention label in the click behavior data; obtaining a second operation weight and a second time-limit weight of the current preset intention label in the search behavior data; obtaining a third operation weight, a third time-limit weight, and an emotional orientation of the current preset intention label in the chat data; and calculating the weight score of the current preference information according to the following formula:

[0007] S=W day1 ×W action1 +W day2 ×W action2 +W day3 ×W action3 ×f

[0008] Where W day1 is the first time-sensitive weight, W day2 is the second time-sensitive weight, W day3 is the third time-sensitive weight, W action1 is the first operation weight, W action2 is the second operation weight, W action3 is the third operation weight, f is the sentiment orientation, and S is the weight score; each preference information is traversed until the weight scores of all preference information are calculated.

[0009] Optionally, the calculation formula of the timeliness weight is:

[0010]

[0011] Where W day is the timeliness weight, and D is the number of days from today.

[0012] Optionally, extracting target preference information from all preference information based on the weight scores includes: classifying the preference information based on the house attributes to which each preference information belongs to obtain multiple preference types; traversing each preference type, comparing the weight scores of each preference information belonging to the same preference type, and taking the preference information with the highest weight score as the target preference information of the current preference type.

[0013] Optionally, the method further includes: obtaining the target preference information of the user at preset time intervals, and storing the target preference information obtained each time into a cache database; extracting the target preference information of the user within a preset time period before the current moment from the cache database, where the length of the preset time period is greater than the preset time interval; performing weighted calculation on the target preference information within the preset time period to generate the second preference information of the user; and pushing housing listings to the user based on the second preference information.

[0014] Optionally, pushing the target property to the user based on the target preference information includes: matching the target preference information with each property information in a property database; if the degree of overlap between the attribute label of the current property information and the preference attribute in the target preference information is above a preset ratio, recommending the current property information to the user.

[0015] According to the second aspect, the present invention provides a housing push device, which includes: a data acquisition module for obtaining the user's click behavior data, search behavior data, chat data and time data in the housing search software, and the time data is used to count the time length of the click behavior data, the search behavior data and the chat data from the current moment; a preference extraction module for extracting all the user's preference information from the click behavior data, the search behavior data and the chat data; a preference analysis module for calculating the weight score of each preference information based on the operation weight and time weight of the click behavior data, the search behavior data and the chat data respectively, the operation weight is the credibility of the click behavior data, the search behavior data and the chat data in representing the user's preference, and the time weight is generated based on the time data, and is used to represent the change in the operation weight over time; a housing push module for extracting target preference information from all the preference information based on the weight score, and pushing target housing to the user according to the target preference information.

[0016] According to the third aspect, an embodiment of the present invention provides a property push device, comprising: a memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the method described in the first aspect or any optional embodiment of the first aspect by executing the computer instructions.

[0017] According to a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to enable the computer to execute the method described in the first aspect or any optional embodiment of the first aspect.

[0018] The technical solution provided by this application has the following advantages:

[0019] The technical solution provided by this application first obtains the click behavior data, search behavior data, chat data and timeliness data in the house-hunting software, and then extracts all the user's preference information from the click behavior data, search behavior data and chat data. The click behavior data, search behavior data and chat data each correspond to different preset operation weights, which are used to measure the credibility of different data types in representing the user's true preferences. The operation weight will also change according to the age of the data. The older the data, the lower the credibility of the preference. Therefore, the timeliness weight is calculated based on the timeliness data to measure the change in the operation weight according to the age of the data. Then, the weight score of each preference information is calculated based on the operation weight and timeliness weight of the click behavior data, search behavior data and chat data. Finally, the target preference information is extracted from all the preference information based on the weight score, and the target housing is pushed to the user based on the target preference information. This achieves the accuracy of recommending housing to users.

[0020] In addition, users' click behaviors include historical records of users browsing, sharing, collecting, booking and signing previews of properties in the house-hunting software. By integrating multiple user operations, the accuracy of determining user target preference information can be further improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The features and advantages of the present invention will be more clearly understood by referring to the accompanying drawings, which are schematic and should not be construed as limiting the present invention in any way. In the accompanying drawings:

[0022] Figure 1 A schematic diagram showing the steps of a method for pushing house listings in one embodiment of the present invention is shown;

[0023] Figure 2 A schematic structural diagram of a housing listing push device according to one embodiment of the present invention is shown;

[0024] Figure 3A schematic structural diagram of a property listing push device in one embodiment of the present invention is shown. DETAILED DESCRIPTION

[0025] To make the purpose, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative efforts shall fall within the scope of protection of the present invention.

[0026] See also Figure 1 In one embodiment, a method for pushing a property listing includes the following steps:

[0027] Step S101: Obtain the user's click behavior data, search behavior data, chat data and timeliness data in the house-hunting software. The timeliness data is used to count the time length from the click behavior data, search behavior data and chat data to the current moment. Specifically, in this embodiment, in addition to obtaining the user's click behavior data, search behavior data and chat data in the house-hunting software, a type of data called timeliness data is also obtained. The timeliness data records the time length from the generation moment of the above three types of behavior data to the current moment. Since the longer the data time, the lower its reference value, the lower the credibility of the user's preference. Therefore, by comprehensively analyzing the user's true preferences through timeliness data and the above three types of behavior data, the accuracy of identifying user preferences can be further improved. In this embodiment, the user's click behavior data is taken from the Shen Ce behavior log, and the user's search behavior data first identifies the user's intended words from the search keywords entered by the user, and then matches the preset intent tags. For the identification of search keywords, a deep learning text recognition model can be used. The specific identification process is a prior art and will not be repeated in the present invention. For user chat data, this embodiment first uses the text classification model TextCNN to identify whether the chat scenario is an apartment-hunting scenario. For example, chat data from a user who has already moved in to consult about daily life issues during their stay is a non-apartment-hunting scenario. A software manager asks the user about their apartment-hunting needs, and the user details their requirements for location, size, and other aspects, which is also a part of the apartment-hunting scenario. Next, the pre-trained model Bert is used for apartment-hunting scenarios to determine whether the current scenario contains a specific preference label, namely a preset intent label. (Since user search terms and chat terms vary, the identified user intent text needs to be standardized to form a preset intent label to facilitate unified analysis and processing of subsequent data. For example, if user A searches for "south" and chats about "south," the standard preset intent label in the software is matched to "facing south.") Furthermore, to improve the accuracy of user intent recognition, this embodiment also uses a vocabulary method to match label words in the chat text. Furthermore, in the question-and-answer dialogue between the manager and the user, the correct intent in the current sentence is determined by combining the previous question and the sentence. The specific process of extracting chat data described above is prior art and will not be repeated here.

[0028] Step S102: Extract all user preference information from click behavior data, search behavior data, and chat data. Specifically, all user preference information is extracted from the three types of behavior data. For example, if a user clicks on a property with a school district, north-facing, and three bedrooms, the corresponding preference information "school district," "north-facing," and "three bedrooms" is extracted. Similarly, all preferences appearing in the user's search behavior data and chat data are extracted for subsequent analysis.

[0029] Step S103: Calculate the weight score of each preference information item based on the operational weight and time-based weight of each of the click behavior data, search behavior data, and chat data. The operational weight represents the credibility of each of the click behavior data, search behavior data, and chat data in representing the user's preference. The time-based weight is generated based on the time-based data and represents the change in the operational weight over time. Specifically, to further improve the accuracy of analyzing the user's true preferences based on the three types of data, operational weights are first preset for the click behavior data, search behavior data, and chat data. In this embodiment, considering that users' interests and preferences generally follow the order of search > chat > click, the preset operational weight ratio also conforms to this order. For example, the operational weight ratio of click behavior data, search behavior data, and chat data is "2:6:5." Furthermore, since the credibility of acquired data decreases with age, this embodiment uses a decay function that decreases over time. The specific value of the time-based weight is obtained by substituting the time length into the decay function. Next, considering the coupling relationship between the four types of data, this embodiment uses multiplication to calculate the final weight score of each preference information item. For example, a certain preference information "facing south" appears in click behavior data, search behavior data, and chat data. The time when "facing south" appears in the three types of data is obtained, and the timeliness weights of "facing south" in the three types of data are calculated to be f1, f2, and f3 respectively. The weight score of "facing south" is 2*f1+5*f2+6*f3.

[0030] Step S104: Extract target preference information from all preference information based on the weight score, and push target properties to the user based on the target preference information. Specifically, in actual situations, a user may generate a lot of preference information, such as "facing south, facing north, facing east, school district, villa, building, 1st floor, 2nd floor, 3rd floor, 3 bedrooms, 2 bedrooms", etc. However, the above preference information is not necessarily the user's true preference information. Therefore, through the weight scores of each preference information obtained in steps S101 to S103, it is possible to know which preference information has high scores, and thus filter out the target preference information with high scores from all the preference information, realizing the function of accurately judging the user's target preference information, and then recommending properties to the user based on the target preference information, greatly improving the accuracy of property recommendations.

[0031] Specifically, in one embodiment, the above step S101 specifically includes the following steps:

[0032] Step 1: Obtain the historical records of users browsing, sharing, collecting, making appointments for viewing, and signing contracts for previewing properties in the house-hunting software, and generate click behavior data based on the preset intention labels marked on the properties in the historical records. Specifically, in this embodiment, the user's click behavior operation data in the house-hunting software is obtained, and the click behavior types include browsing, sharing, collecting, making appointments for viewing, and signing contracts for previewing, so that the source types of user click behavior data are more diverse, thereby improving the accuracy of subsequent analysis. In addition, this embodiment subdivides the operation weights of the click behavior data into each click operation type, so that the calculation of the weight score is more accurate. For example: this embodiment presets the operation weight ratios of browsing, sharing, collecting, making appointments for viewing, and signing contracts for previewing to be 1:1.3:1.8:2.4:4 respectively. Correspondingly, if the preference information "facing south" is browsed and shared respectively, then when calculating the weight score, "facing south" needs to be calculated twice with weights of 1 and 1.3 respectively. The click behavior data directly uses the standardized preset intention labels marked on the property information in the software, so no conversion is required.

[0033] Step 2: Convert the user's search terms in the house-hunting software into preset intent tags, and generate search behavior data based on the converted preset intent tags.

[0034] Step 3: Extract the intent keywords and the sentiment orientation of the intent keywords from the chat records of users in the house-hunting software, and convert the intent keywords into preset intent tags with sentiment orientation. Then, generate chat data based on the converted preset intent tags. The sentiment orientation is used to determine whether the intent keywords are the user's true preference.

[0035] Specifically, the specific method of obtaining the search behavior data and chat data refers to step S101 and will not be repeated here. It should be noted that the intent label of the chat data in this embodiment is emotionally oriented, thereby improving the accuracy of judging the user's true preference. For example: in this embodiment, the positive emotion is set to +1, and the negative emotion is set to -1. Assuming that the user's dialogue corpus is "I want to find a house facing south, especially not facing east", the two preset intent labels in the corpus are identified as "facing south" and "facing east", and the two preset intent labels are added with emotional orientations of +1 and -1 respectively. In the subsequent weight score calculation process, the weight score calculation of "facing east" is involved, and the calculation operation of the chat data part is subtraction, thereby further improving the calculation accuracy of the weight score.

[0036] Obtain the time length from the moment when the click behavior data, search behavior data, and chat data occurred to the current moment, and generate time-sensitive data based on the time length.

[0037] Specifically, in one embodiment, the above step S103 specifically includes the following steps:

[0038] Step 4: Get the current preset intent label corresponding to the current preference information.

[0039] Step 5: Obtain the first operation weight and first timeliness weight of the current preset intent tag in the click behavior data.

[0040] Step 6: Obtain the second operation weight and second timeliness weight of the current preset intent tag in the search behavior data.

[0041] Step 7: Obtain the third operation weight, third timeliness weight, and emotional orientation of the current preset intent tag in the chat data.

[0042] Step 8: Calculate the weight score of the current preference information according to the following formula:

[0043] S=W day1 ×W action1 +W day2 ×W action2 +W day3 ×W action3 ×f

[0044] Where W day1 is the first time-sensitive weight, W day2 is the second time-sensitive weight, W day3 is the third time-sensitive weight, W action1 is the first operation weight, W action2 is the second operation weight, W action3 is the third operation weight, f is the sentiment orientation, and S is the weight score;

[0045] Step 9: Traverse each preference information until the weight scores of all preference information are calculated.

[0046] Specifically, let's continue with the examples in steps one to three. For example: first, extract the current preset intention label corresponding to the current preference information from the user's various preference information. In this embodiment, "facing south" is extracted. After judging that "facing south" appears in all three types of data, the operation weight ratios of the three types of data are 2:6:5, respectively. The time weight is calculated to be 0.5, 0.8, and 0.7. In the chat data, the emotional orientation is +1, then the weight score of "facing south" is calculated to be 2*0.5+6*0.8+5*0.7*1=9.3. Similarly, traverse the user's various preference information until the weight scores of all preference information are calculated. Through the above steps, the user preference information appearing in each type of data is accurately integrated together for analysis, which greatly improves the accuracy of identifying the user's true preferences. Specifically, in this embodiment, the calculation formula for the timeliness weight is:

[0047]

[0048] Where W day is the timeliness weight, and D is the number of days since today. Timeliness weight is calculated in days. The longer the distance between the data collection date and the day before yesterday, the smaller the weight and the lower the credibility of the data.

[0049] Specifically, in one embodiment, the above step S104 includes the following steps:

[0050] Step 10: Classify the preference information based on the property attributes to which each preference information belongs, and obtain multiple preference types.

[0051] Step 11: traverse each preference type, compare the weight scores of each preference information belonging to the same preference type, and use the preference information with the highest weight score as the target preference information of the current preference type.

[0052] Specifically, to determine a user's true preferences, target preference information with high weight scores is extracted from all of the user's preference information. However, the various preference information correspond to different property attributes, and filtering based directly on scores is inevitably inaccurate. Therefore, in this embodiment, preference information is first classified based on the property attribute to which each preference information belongs, resulting in multiple preference types. For example, "South-facing," "East-facing," and "North-facing" all belong to the property orientation attribute, while "Three-bedroom," "Two-bedroom," and "One-bedroom" belong to the property number attribute. Preference information belonging to the same attribute is grouped together to obtain multiple preference types. Then, for each preference type, weight scores are compared within the preference type. For example, if "South-facing," "East-facing," and "North-facing" all appear in the user's three behavioral data categories, but the weight scores are 9.3, 8.2, and 7.1, respectively, the preference information with the highest score is selected as the target preference information within that type, namely, "South-facing." Subsequently, each preference type is traversed to obtain the user's complete target preference information, thereby accurately determining the user's desired property orientation, area, number of rooms, location, and so on, significantly improving the accuracy of property recommendations based on the target preference information.

[0053] Specifically, in one embodiment, a method for pushing house listings provided by an embodiment of the present invention further includes the following steps:

[0054] Step 12: Obtain the user's target preference information at preset time intervals, and store the target preference information obtained each time into a cache database.

[0055] Step 13: extracting the user's target preference information within a preset time period before the current moment from the cache database, where the length of the preset time period is greater than the preset time interval;

[0056] The target preference information within a preset time period is weightedly calculated to generate the user's second preference information.

[0057] Step 14: Push properties to the user based on the second preference information.

[0058] Specifically, in this embodiment, the user's target preference information is periodically extracted and stored in a cache database (the embodiment of the present invention uses a Redis database for target preference information caching). When a property is needed for a user, the target cache information for a period of time is retrieved from the cache database. The retrieved period is longer than the extraction period, ensuring that the stored target preference information is retrieved from the cache database at least twice. The multiple target preference information retrieved is then weighted (the embodiment of the present invention uses average weighting, but the present invention is not limited to this), thereby further obtaining a second preference information that more accurately represents the user's true preferences. This further improves the accuracy of subsequent recommendations for suitable properties for the user.

[0059] Specifically, in one embodiment, the above step S104 includes the following steps:

[0060] Step 15: Match the target preference information with each property listing in the property database.

[0061] Step 16: If the degree of overlap between the attribute labels of the current property information and the preferred attributes in the target preference information is above a preset ratio, the current property information is recommended to the user.

[0062] Specifically, after obtaining the user's target preference information, the user's target preference information is matched with each property information in the property database. For example, a property information has 10 attribute tags, 9 of which can find corresponding information in the target preference information, and the matching overlap reaches 90%. In this embodiment, property information with a matching degree of more than 70% is considered to be matched, and the above property information is pushed to the user. After traversing all property information, all property information that meets the preset ratio conditions is recommended to the user. The property recommendation method provided by the present invention is applied to scenarios such as housekeeper recommendation, recommendation system, search system, user portrait, etc., so that users can accurately select their favorite houses.

[0063] Through the above steps, the technical solution provided by this application first obtains the click behavior data, search behavior data, chat data and timeliness data in the house-hunting software, and then extracts all the user's preference information from the click behavior data, search behavior data and chat data, wherein the click behavior data, search behavior data and chat data each correspond to a different preset operation weight, which is used to measure the credibility of different data types in representing the user's true preference. The operation weight will also change according to the age of the data. The older the data, the lower the credibility of the preference. Therefore, the timeliness weight is calculated based on the timeliness data to measure the change in the operation weight according to the age of the data. Then, the weight score of each preference information is calculated based on the operation weight and timeliness weight of the click behavior data, search behavior data and chat data. Finally, the target preference information is extracted from all the preference information based on the weight score, and the target house is pushed to the user based on the target preference information. This achieves the accuracy of recommending houses to users.

[0064] In addition, users' click behaviors include historical records of users browsing, sharing, collecting, booking and signing previews of properties in the house-hunting software. By integrating multiple user operations, the accuracy of determining user target preference information can be further improved.

[0065] like Figure 2 As shown, this embodiment also provides a housing listing push device, which includes:

[0066] Data collection module 101 is used to obtain user click behavior data, search behavior data, chat data, and time-sensitive data in the house-hunting software. Time-sensitive data is used to calculate the length of time from the current moment to the current moment in the click behavior data, search behavior data, and chat data. For details, please refer to the description of step S101 in the above method embodiment and will not be repeated here.

[0067] The preference extraction module 102 is used to extract all the user's preference information from the click behavior data, search behavior data and chat data. For details, please refer to the relevant description of step S102 in the above method embodiment, which will not be repeated here.

[0068] Preference analysis module 103 is configured to calculate weight scores for each preference information based on the operational weights and time-based weights of the click behavior data, search behavior data, and chat data. The operational weights represent the credibility of the click behavior data, search behavior data, and chat data in representing the user's preference. The time-based weights are generated based on the time-based data and represent the amount by which the operational weights increase over time. For details, see the description of step S103 in the above method embodiment and will not be repeated here.

[0069] The property push module 104 is used to extract target preference information from all preference information based on the weight score and push target properties to the user based on the target preference information. For details, please refer to the description of step S104 in the above method embodiment, which will not be repeated here.

[0070] The housing resource push device provided in an embodiment of the present invention is used to execute the housing resource push method provided in the above embodiment. Its implementation method and principle are the same. For details, please refer to the relevant description of the above method embodiment and will not be repeated here.

[0071] Through the collaborative efforts of the aforementioned components, the system first obtains click behavior data, search behavior data, chat data, and timeliness data from the house-hunting software. It then extracts the user's complete preference information from these data. Each of these data types is assigned a preset action weight, which measures the credibility of each data type in representing the user's true preferences. The action weight also changes based on the age of the data; older data has lower preference credibility. Based on the timeliness data, a timeliness weight is calculated to measure how the action weight changes over time. Then, based on the action weights and timeliness weights of each of the click behavior data, search behavior data, and chat data, a weight score is calculated for each preference information item. Finally, based on the weight score, the target preference information is extracted from the total preference information, and the target listing is pushed to the user based on the target preference information. This ensures accurate house recommendations for users.

[0072] In addition, users' click behaviors include historical records of users browsing, sharing, collecting, booking and signing previews of properties in the house-hunting software. By integrating multiple user operations, the accuracy of determining user target preference information can be further improved.

[0073] Figure 3 A property push device according to an embodiment of the present invention is shown. The device includes a processor 901 and a memory 902, which can be connected via a bus or other means. Figure 3 The bus connection is taken as an example.

[0074] The processor 901 may be a central processing unit (CPU). The processor 901 may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, or a combination of the above chips.

[0075] Memory 902, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs, non-transitory computer executable programs, and modules, such as the program instructions / modules corresponding to the methods in the above-described method embodiments. Processor 901 executes the non-transitory software programs, instructions, and modules stored in memory 902 to perform various processor functions and data processing, thereby implementing the methods in the above-described method embodiments.

[0076] The memory 902 may include a program storage area and a data storage area, wherein the program storage area may store an operating system and application programs required for at least one function; the data storage area may store data created by the processor 901, etc. In addition, the memory 902 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. In some embodiments, the memory 902 may optionally include a memory remotely located relative to the processor 901, and these remote memories may be connected to the processor 901 via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0077] One or more modules are stored in the memory 902 and, when executed by the processor 901 , perform the method in the above method embodiment.

[0078] The specific details of the above-mentioned house push device can be understood by referring to the corresponding descriptions and effects in the above-mentioned method embodiment, and will not be repeated here.

[0079] Those skilled in the art will appreciate that all or part of the processes in the above-described embodiments can be implemented by instructing related hardware through a computer program. The implemented program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the above-described embodiments. The storage medium can be a magnetic disk, an optical disk, a read-only memory (ROM), a random access memory (RAM), a flash memory, a hard disk drive (HDD), or a solid-state drive (SSD). The storage medium can also include a combination of the above-mentioned types of memory.

[0080] Although the embodiments of the present invention have been described with reference to the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention. Such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A method for pushing house listings, characterized in that: The method comprises: Obtaining the user's click behavior data, search behavior data, chat data and timeliness data in the house-hunting software, wherein the timeliness data is used to count the time length between the click behavior data, the search behavior data and the chat data and the current moment; obtaining the user's click behavior data, search behavior data, chat data and timeliness data includes: obtaining the user's historical records of browsing, sharing, collecting, booking and signing preview properties in the house-hunting software, and generating the click behavior data based on the preset intention tags marked with the properties in the historical records; converting the user's search terms in the house-hunting software into preset intention tags, and generating the search behavior data based on the converted preset intention tags; extracting the intent keywords and the emotional orientation of the intent keywords in the chat records of the user in the house-hunting software, and converting the intent keywords into preset intention tags with emotional orientation, and then generating the chat data based on the converted preset intention tags, wherein the emotional orientation is used to determine whether the intent keywords are the user's true preference; obtaining the time length from the moment when the click behavior data, the search behavior data and the chat data each behavior occurs to the current moment, and generating the timeliness data based on the time length; Extracting all preference information of the user from the click behavior data, search behavior data and chat data; Calculating a weight score for each preference information based on the operation weight and time-sensitive weight of each of the click behavior data, search behavior data, and chat data, wherein the operation weight is the credibility of each of the click behavior data, search behavior data, and chat data in representing the user's preference; and the time-sensitive weight is generated based on the time-sensitive data and is used to represent the amount by which the operation weight increases over time; Target preference information is extracted from all the preference information based on the weight scores, and target housing is pushed to the user based on the target preference information; extracting the target preference information from all the preference information based on the weight scores includes: classifying the preference information based on the housing attributes to which each piece of preference information belongs to obtain multiple preference types; traversing each preference type, comparing the weight scores of each piece of preference information belonging to the same preference type, and taking the preference information with the highest weight score as the target preference information of the current preference type.

2. The method according to claim 1, characterized in that The calculating of the weight scores of the respective preference information based on the operation weights and timeliness weights of the click behavior data, search behavior data, and chat data includes: Get the current preset intent label corresponding to the current preference information; Obtaining a first operation weight and a first timeliness weight of the current preset intention tag in the click behavior data; Obtaining a second operation weight and a second timeliness weight of the current preset intention tag in the search behavior data; Obtaining a third operation weight, a third timeliness weight, and an emotional orientation of the current preset intent tag in the chat data; The weight score of the current preference information is calculated according to the following formula: Where, is the first time-sensitive weight, is the second time-sensitive weight, is the third time weight, is the first operation weight, is the second operation weight, is the third operation weight, f It is emotion-oriented. S is the weight score; Traverse each preference information until the weight scores of all preference information are calculated.

3. The method according to claim 2, characterized in that The calculation formula of the timeliness weight is: Where, is the timeliness weight, and D is the number of days from today.

4. The method according to claim 1, wherein The method further comprises: Acquire the target preference information of the user at preset time intervals, and store the acquired target preference information into a cache database; extracting target preference information of the user within a preset time period before the current moment from the cache database, where the length of the preset time period is greater than the preset time interval; Performing weighted calculation on the target preference information within the preset time period to generate second preference information of the user; Pushing housing listings to the user according to the second preference information.

5. The method according to claim 1, wherein Pushing target properties to the user according to the target preference information includes: Matching the target preference information with each property listing in a property listing database; If the degree of overlap between the attribute labels of the current property information and the preference attributes in the target preference information is above a preset ratio, the current property information is recommended to the user.

6. A property push device, characterized in that: The device comprises: A data acquisition module is used to obtain the user's click behavior data, search behavior data, chat data and timeliness data in the house-hunting software, wherein the timeliness data is used to count the time length between the click behavior data, the search behavior data and the chat data and the current moment; the acquisition of the user's click behavior data, search behavior data, chat data and timeliness data includes: obtaining the user's historical records of browsing, sharing, collecting, booking and signing preview properties in the house-hunting software, and generating the click behavior data based on the preset intention tags marked with the properties in the historical records; converting the user's search terms in the house-hunting software into preset intention tags, and generating the search behavior data based on the converted preset intention tags; extracting the intent keywords and the emotional orientation of the intent keywords in the chat records of the user in the house-hunting software, and converting the intent keywords into preset intention tags with emotional orientation, and then generating the chat data based on the converted preset intention tags, wherein the emotional orientation is used to determine whether the intent keywords are the user's true preference; obtaining the time length from the time when the click behavior data, the search behavior data and the chat data each behavior occurs to the current moment, and generating the timeliness data based on the time length; A preference extraction module is used to extract all the user's preference information from the click behavior data, search behavior data and chat data; a preference analysis module for calculating a weight score for each piece of preference information based on the operation weight and time-sensitive weight of each of the click behavior data, search behavior data, and chat data, wherein the operation weight is the credibility of each of the click behavior data, search behavior data, and chat data in representing the user's preference; and the time-sensitive weight is generated based on the time-sensitive data and is used to represent the amount by which the operation weight increases over time; A housing push module is used to extract target preference information from all the preference information based on the weight score, and push target housing to the user according to the target preference information; the extraction of target preference information from all the preference information based on the weight score includes: classifying the preference information based on the housing attributes to which each piece of preference information belongs to obtain multiple preference types; traversing each preference type, comparing the weight scores of each piece of preference information belonging to the same preference type, and taking the preference information with the highest weight score as the target preference information of the current preference type.

7. A property push device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the method according to any one of claims 1 to 5 by executing the computer instructions.

8. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable the computer to execute the method according to any one of claims 1 to 5.

Images