Recommendation sorting method based on generation of user interest labels

Abstract

The invention discloses a recommendation sorting method based on generation of user interest labels, and the method comprises the following steps: S1, setting label weights according to different labels; s2, obtaining and processing the demand description of the user through the online user action; s3, the expected location of the user is confirmed by the online behavior of the user, and the weight is reduced based on the occurrence time of the user action; s4, according to historical screening and searching behaviors of the users, collecting user tags, and matching the inquired house attributes with the collected user tags; if the matching succeeds, the value is 1; if the matching is unsuccessful, the value is 0; s5, calculating the sorting score of the housing resources, wherein the sorting score is equal to sigma label * matching degree * label weight * time weight; and S6, sorting the housing resources according to the sorting score. The method comprises the following steps: collecting and classifying user interest data, and calculating the weight of each latitude label; the weekly-dimension CTR click efficiency of the recommendation type list is improved by 35%, and the search result list is level.

Description

Technical Field

[0001] The present invention relates to the technical field of data analysis applications, and in particular to a recommendation ranking method based on generating user interest tags. Background Art

[0002] With the development of the Internet, more and more services can be obtained through the Internet. Among them, the services obtained through the Internet may include online house transactions. Users can browse the house information posted by real estate developers or landlords on websites that provide online house transaction services, contact real estate agents or landlords on the website, and conduct house transactions on the Internet. Generally speaking, when users use websites that provide online house transaction services, they need to find house information that meets their needs on the website. Therefore, users can use the search service provided by the website to search and filter house information by entering keywords or selecting several categories from preset categories. Generally speaking, the website can display the user's house information search results in a list format.

[0003] Currently, many C-end apps, such as the home page, channel page, and search page request without a location, do not have personalized reflection of user demand information, and the ranking of houses that meet user demands is often not very high. The problems mainly manifest in the following two aspects: low relevance: I was looking for houses in Pudong before, but every time I open the home page, I still see houses in Puxi; low effectiveness: I just found a 1-bedroom house in Pudong, but I return to the home page to continue looking for 2-bedroom houses in other districts.

[0004] In summary, the above problems will cause users to give up using the homepage, reduce the homepage’s positioning as a testing ground for user needs and a module for gathering needs, and reduce the overall user experience. Summary of the invention

[0005] The purpose of the present invention is to provide a recommendation ranking method based on generating user interest tags to solve the problems in the prior art.

[0006] To achieve the above object, the present invention provides the following technical solution: a recommendation ranking method based on generating user interest tags, comprising the following steps: S1. Set the label weight according to different labels; S2, obtain and process user demand descriptions through online user actions; S3, the user's expected location is determined by the user's online behavior, and the weight is reduced based on the time when the user's action occurs; S4. Collect user tags based on the user's historical search behavior, and match the queried house attributes with the collected user tags; if the match is successful, the value is 1; if the match is unsuccessful, the value is 0; S5. Calculate the ranking score of the house, ranking score = ∑ label * matching degree * label weight * time weight; S6. Sort the properties according to the ranking points.

[0007] Preferably, the tags include neighborhoods of interest, subway stations of interest, business districts of interest, administrative districts of interest, house types of interest, price ranges of interest, and house types.

[0008] Preferably, the label weight of the community where the house is located is 30; the label weight of the business district where the house is located is 10; the label weight of the administrative district where the house is located is 5; the label weight of the number of rooms is 1; the label weight of the rental price is 3; the label weight of the commuting node is 50; the label weight of the check-in time is 10; the label weight of the label with video is 1; the label weight of the label with VR is 1; the label weight of the label with pictures is 200; and the label weight of the house type is 10.

[0009] Preferably, the coefficient of the time weight=(7th root of 0.5)^((current timestamp-most recent timestamp that satisfies the condition) / (24*60*60*1000)).

[0010] Compared with the prior art, the beneficial effects of the present invention are: collecting and classifying user interest data, further proposing a weight calculation scheme for each dimensional label, and applying it to different business scenarios - list sorting; the CTR click efficiency of the recommended type list (home page, no fewer results) in the weekly dimension is increased by 35%, while the search result list remains the same. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings: Figure 1 is a flow chart of the present invention; Figure 2 It is a structural schematic diagram of a calculation example of the present invention; Figure 3 It is a time sequence diagram of collecting user behaviors of the present invention; Figure 4 It is a search sorting timing diagram of the present invention; Figure 5 It is a weekly dimension CTR graph of the recommendation type list of the present invention. DETAILED DESCRIPTION

[0012] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, 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 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 ordinary technicians in this field without creative work are within the scope of protection of the present invention. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention.

[0013] See also Figure 1 , 3 , 4 and 5, in an embodiment of the present invention, a recommendation ranking method based on generating user interest tags comprises the following steps: S1. Set the label weight according to different labels; S2, obtain and process user demand descriptions through online user actions; S3, the user's expected location is determined by the user's online behavior, and the weight is reduced based on the time when the user's action occurs; S4. According to the user's historical search behavior, user tags are collected first, and the queried house attributes are matched with the collected user tags. If the match is successful, the value is 1; if the match is unsuccessful, the value is 0. (According to the user's historical search behavior, user tags are collected first, that is, the query conditions for querying houses, such as administrative districts, business districts, subway stations, and living rooms are collected; after querying the house results, the attribute values ​​of each house (the administrative district, business district, subway station, and living room where the house is located) are matched with the user tags, 0 and 1 represent whether the match is hit or not, and if the match is equal, the value is 1, which means hit, otherwise no hit.) S5. Calculate the ranking score of the house, ranking score = ∑ label * matching degree * label weight * time weight; S6. Sort the properties according to the ranking points.

[0014] Preferably, the tags include neighborhoods of interest, subway stations of interest, business districts of interest, administrative districts of interest, house types of interest, price ranges of interest, and house types.

[0015] Preferably, the label weight of the community where the house is located is 30; the label weight of the business district where the house is located is 10; the label weight of the administrative district where the house is located is 5; the label weight of the number of rooms is 1; the label weight of the rental price is 3; the label weight of the commuting node is 50; the label weight of the check-in time is 10; the label weight of the label with video is 1; the label weight of the label with VR is 1; the label weight of the label with pictures is 200; and the label weight of the house type is 10.

[0016] Preferably, the coefficient of time weight = (7th root of 0.5)^((current timestamp - most recently satisfied timestamp) / (24*60*60*1000)). If the most recently satisfied timestamp (i.e., the time when the user tag was collected) is today, the calculation result is close to 1. If the satisfied timestamp is yesterday, the calculation result is smaller and the score is smaller. The tags that the user has recently searched for are considered to be more relevant and have higher scores. When sorting batches of houses, the current house is ranked higher.

[0017] The ranking directly reflects the user’s actual housing search needs; The user's expected location is determined by the user's online behavior (browsing listings, browsing neighborhoods, searching, filling in information, and adding listings and neighborhoods to favorites), and the weight is decreased based on the time when the user's action occurred; The user's demand for housing is reflected in real time in the user's subsequent sorting of housing sources; Disassembly and storage of key fields of user requirements; Obtain and process user demand descriptions through online user actions; Understanding of user demand fields is used as the information source of user CDP and is provided for use by other businesses; The matching degree of the region is added as location relevance in the sorting; The user's expected location is determined by the user's online behavior (browsing listings, browsing neighborhoods, searching, filling in information, and adding listings and neighborhoods to favorites), and the weight is decreased based on the time when the user's action occurred; User tag collection methods: User page filtering Location: subway, business district, administrative district; Rental prices; House type: house type, rooms; New user onboarding The user selects the desired apartment type: house type, rooms; 3. Listings clicked by users All labels will be collected, including the neighborhood (neighborhood label), location (business district, subway, administrative district), price, etc. For example, Figure 2 ; Set its label weight to match the neighborhood of interest, subway station of interest, and price range of interest; match the queried house attributes with the collected user labels; if the match is successful, the value is 1; if the match is unsuccessful, the value is 0; Property 1 is located in Community A, at Metro Station A and costs RMB 1,500, so the calculation is 1*0.9+1*0.9+1*0.8=2.6.

[0018] Property 2 is located in Community B, at Metro Station C and costs RMB 3,500, which is calculated as 1*0.8=0.8.

[0019] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A recommendation ranking method based on generating user interest tags, characterized in that: The following steps are involved: S1. Set the label weight according to different labels; S2, obtain and process user demand descriptions through online user actions; S3, the user's expected location is determined by the user's online behavior, and the weight is reduced based on the time when the user's action occurs; S4. Collect user tags based on the user's historical search behavior, and match the queried house attributes with the collected user tags; if the match is successful, the value is 1; if the match is unsuccessful, the value is 0; S5. Calculate the ranking score of the house, ranking score = ∑ label * matching degree * label weight * time weight; S6. Sort the properties according to the ranking points.

2. According to claim 1, a recommendation ranking method based on generating user interest tags is characterized in that: The tags include interest communities, interest subway stations, interest business districts, interest administrative districts, interest house types, interest price ranges, and house types.

3. A recommendation ranking method based on generating user interest tags according to claim 1, characterized in that: The label weight of the neighborhood where the property is located is 30; the label weight of the business district where the property is located is 10; the label weight of the administrative district where the property is located is 5; the label weight of the number of rooms is 1; the label weight of the rental price is 3; the label weight of the commuting node is 50; the label weight of the check-in time is 10; the label weight of the label with video is 1; the label weight of the label with VR is 1; the label weight of the label with pictures is 200; the label weight of the house type is 10.

4. The recommendation ranking method based on generating user interest tags according to claim 1, characterized in that: The time weight coefficient = (7th root of 0.5)^((current timestamp - most recent timestamp that meets the condition) / (24*60*60*1000)).

Images