CN112559893A - User recommendation method in internet matching social scene - Google Patents
User recommendation method in internet matching social scene Download PDFInfo
- Publication number
- CN112559893A CN112559893A CN202110030756.0A CN202110030756A CN112559893A CN 112559893 A CN112559893 A CN 112559893A CN 202110030756 A CN202110030756 A CN 202110030756A CN 112559893 A CN112559893 A CN 112559893A
- Authority
- CN
- China
- Prior art keywords
- user
- pool
- recommendation
- users
- current
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9536—Search customisation based on social or collaborative filtering
Landscapes
- Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention relates to a user recommendation method in an internet matching social scene, which comprises user A level division, user recommendation pool division, layered extraction of a user B in a user recommendation pool, user A preference behavior data collection, user A level division correction according to the behavior data of the user A and user B recommendation in the user recommendation pool. The method extracts the user B from different recommendation pools and recommends the user B to the user A, tracks and collects the preference of the user A on the recommended user B and the preference of the user B on the user A, and corrects the hierarchy division of the user A and the preference of the user B recommended to the user A, so that the user B which is possibly interested in the user A is recommended to the user A, and the matching rate and the use experience are improved.
Description
Technical Field
The invention relates to the technical field of computers, in particular to a user recommendation method in an internet matching social scene.
Background
At present, the pace of life is continuously accelerated, social networking services are popular, and a common mode of social networking services is a matching mode, that is, a user publishes media such as characters, photos, music and videos or browses media such as photos, characters, music and videos published by other users on social software, and both parties express interests of the other party by praise or some operations, and can perform social behaviors such as chatting or video telephony.
The main problem to be solved by the internet social recommendation algorithm is that users who may be interested in recommending users accurately. Common algorithms such as a collaborative filtering algorithm, a similarity recommendation algorithm and the like have the core idea that users are portrayed through user issued information and praise behaviors, and users similar to object features of interest of user history are found out.
Considering that the social behavior is a bidirectional selection process, on one hand, a user needs to have new matching to maintain social motivation of the user, and on the other hand, the user needs to search for an object really fit with the user; the requirement cannot be met by recommending users with similar hobbies, ages, distances and other similar dimensions to the users at a glance.
And the interest points of people are different from the interest points of people for commodities, most of the interest points of people are multidirectional and uncertain, a user may have interest in various types of people, and the user may not know the interest of the user per se, so that a single E-commerce recommendation algorithm cannot enable the user to have better social experience.
Disclosure of Invention
In order to solve the technical problems, the invention provides a user recommendation method in an internet matching social scene, which extracts users from a plurality of recommendation pools, meets the matching requirements of the users on one hand, recommends new matching users for the users on the other hand, and improves the user experience, and the technical purpose of the invention is realized by the following technical scheme:
a user recommendation method in an internet matching social scene comprises user A level division, user recommendation pool division, user B in a user recommendation pool is extracted in a layered mode, user A preference behavior data collection, user A level division correction according to the user A behavior data, and the user B extracted from the user recommendation pool.
Further, when the user A is divided into levels, the grades are divided according to the grade of the user A, and the grade of the user A comprises a basic grade b and a floating grade a; the basic score b is obtained by weighted calculation according to the basic information of the user A, and the floating score a is obtained by weighted calculation according to the times that the user A is preferred by other users/the times that the user A recommends to other users.
Further, the user recommendation pool comprises a one-way relation pool, a current level head user pool and an individual recommendation pool, and the individual recommendation pool is formed by matching the labels of the user A and the label of the user B; the unidirectional relation pool is generated according to the preference behavior of the user B to the user A; and the current hierarchy head user pool is formed according to the hierarchy division of the user A, and the hierarchy of the user B in the current hierarchy head user pool is the same as that of the user A.
Further, the priorities of the personalized recommendation pool, the one-way relationship pool and the current hierarchy head user pool are sequentially reduced, and when the user B in the user recommendation pool is extracted in a hierarchical mode, the user B is sequentially extracted from the personalized recommendation pool, the one-way relationship pool and the current hierarchy head user pool.
Further, the extraction ratio of the personalized recommendation pool is greater than that of the one-way relationship pool and the current-level head user pool, and the extraction ratio of the one-way relationship pool is greater than that of the current-level head user pool.
Further, when the user B is hierarchically extracted from the user recommendation pool, the previous priority extraction shortage is extracted from the next priority to supplement the extraction proportion.
Further, when the user B is extracted from the personalized recommendation pool, the one-way relationship pool and the current level head user pool, ranking the user B extracted from the personalized recommendation pool, the one-way relationship pool and the current level head user pool respectively, and extracting the user B with the top rank.
Further, the user B is a user set recommended to the user a, and the user a itself is not included in the user set.
Compared with the prior art, the method has the advantages that the user B is extracted from the three recommendation pools and recommended to the user A, the preference of the user A on the recommended user B and the preference of the user B on the user A are tracked and collected, the hierarchy division of the user A and the user B recommended to the user A are corrected, on one hand, the user B is recommended to the user A from different angles, more accurate matching requirements can be met by correcting the hierarchy division of the user A and the user B recommended to the user A, and the matching rate and the use experience are improved; on the other hand, the user A can be provided with a continuous user B through the recommendation pool.
Drawings
FIG. 1 is a flow chart of a user recommendation pool composition and recommendation algorithm in the present invention.
Detailed Description
The technical solution of the present invention is further described below with reference to specific embodiments:
a user recommendation method in an internet matching social scene comprises user A hierarchy division, user recommendation pool division, user B in a user recommendation pool is extracted in a hierarchical mode, user A behavior data collection is carried out, user A self hierarchy division is corrected according to the behavior data of the user A, and the user B recommended to the user A is obtained. The source of the user B is a plurality of users A, a plurality of other users A except the user A are used as the users B recommended to the user A, and each user A can be used as one of the users B recommended to other users.
Specifically, each user A fills in basic data during registration, such as school, occupation, age, height, whether the basic data passes real person verification, issued dynamic information and other basic information, and calculates the basic score of each user A, and then the basic score is calculated as b; the number of times (exposure value) that each user a recommends to other users, and the number of times/exposure value that each user a prefers to other users after recommending to other users is counted as a floating score of the user a, for example, the user a recommends to other users 10 times, where the number of times preferred by other users is 3, the floating score is 3/10=0.3, the preference may be liked, commented, browsed by other users, and the total score of the user is calculated from the basic score and the floating score. For example, the total score y = ax + b, where x is a floating coefficient, the total score y of the user a is obtained through calculation, and the user a is classified into levels according to the total score. For example, the total score is divided into 100 points, and divided into four levels according to 0-25, 26-50, 51-75 and 76-100, the total score of the user A is divided into corresponding levels in corresponding level intervals, the level division is not fixed, and changes can be generated according to the change of the floating score a. The purpose of hierarchical division is to initially screen out users with the same level as that of the current user A from all other users, and to be more beneficial to picking out groups which are similar to the user A in the same level, wherein the groups are similar in appearance, academic history, occupation and the like, and the groups which are similar are more in line with circle-level social contact, so that the recommendation efficiency is greatly improved.
The user recommendation pool is a data source recommended to the user A and is composed of a plurality of other users A, and the users in the user recommendation pool become the user B. The recommendation pool is divided into three parts, namely a personalized recommendation pool, a one-way relation pool and a current hierarchy head user pool, the hierarchy evaluation mode of a user B recommended to the user A is the same as that of the user A, and the hierarchy of the user B recommended to the user A is the same as that of the user A.
The personalized recommendation pool is formed by matching the similarity of a user A label and a user B label, the information of each user and preference data of contents such as photos, characters, music and the like in the using process are recorded, the user label is formed according to the preference record, similarity calculation is carried out according to the user label to find out a similar user group, weight calculation is carried out on the selected group, the calculation dimensions such as the similarity of published contents, the liveness of the user, the score of the user, the geographic position of the user and the like are calculated, and the weight coefficient of the similarity of the published contents is highest. For example, the label of the currently issued content of the user a is "shanghai", "cat" or "game", the user B containing the label is inquired, the label similarity is calculated, dimension weight calculation such as the issued content, the user activity, the user score, the user geographical position distance and the like is performed on the screened user B, the data with the top weight calculation result is selected, for example, 6 data are extracted, and the data with the top weight calculation result ranked six are extracted and recommended to the user a.
The unidirectional relation pool is generated according to the preference behavior of the user B to the user A, and the recommendation aims to enable the user A to see the user B interested in the user A, namely, a plurality of users B possibly interested in the user A are provided for the user A, and the user A is enabled to select the user B possibly interested in the user A from the group interested in the user A, so that continuous matching experience is obtained, and the difficulty of no feedback in social contact is eliminated.
The recommendation logic of the unidirectional relation pool is calculated after weighting according to three dimensions of the geographic position, the total score of the user, the label matching degree of the information such as the photo published by the user and the like, wherein the weight of the geographic position distance is the highest, such as:
calculating a recommendation weight score: and (2) recimend _ score = a distance + B score + c tags, wherein a is a distance weight coefficient, distance is the distance difference between the user A and the user B, B is a user score coefficient, score is the score difference between the user A and the user B, c is a user tag coefficient, and tags are the tag similarity of the user A and the user B.
For example, 3 users B need to be extracted from the unidirectional relation pool and recommended to the user a, and the user B with the top three of the weight calculation result is extracted and recommended to the user a.
The head user pool of the current level is composed of users B of which the current level is the same as that of the users A, the users represent groups similar to the ages, the professions and the like of the users A, the dimension weighted by the recommendation coefficients of the users comprises the total user score and the geographical position distance of the users, and the total user score is highest in weight. And assuming that 1 user B needs to be extracted from the current hierarchy user pool and recommended to the user A, extracting the user with the first ranking of the weight calculation result.
The priorities of the personalized recommendation pool, the one-way relation pool and the current hierarchy head user pool are sequentially reduced, namely the user B is extracted from the personalized recommendation pool, then the one-way relation pool and finally the current hierarchy head user pool; during extraction, the extraction proportion of the personalized recommendation pool is greater than that of the one-way relation pool, and the extraction proportion of the one-way relation pool is greater than that of the current level head user pool, for example, the personalized recommendation pool, the one-way relation pool and the current level head user pool are respectively extracted according to the ratio of 6:3: 1; when the data source in the user recommendation pool of the current priority is not enough to meet the extraction ratio, extracting data from the next priority for complementing, for example, a newly registered user, the unidirectional relationship pool of which may be empty, extracting data from the head user pool of the current hierarchy for complementing the extraction ratio of the unidirectional relationship pool, specifically, if extracting 6 from the personalized recommendation pool, extracting 3 from the unidirectional relationship pool, only extracting 1 from the head user pool of the current hierarchy, but only 1 data source in the unidirectional relationship pool, extracting 2 from the head user pool of the current hierarchy for complementing the extraction ratio of the unidirectional relationship pool, and then extracting 1 from the head user pool of the current hierarchy, wherein the actual extraction ratio is the personalized recommendation pool: the one-way relationship pool: current hierarchy head user pool =6:1: 3.
Because the data sources in the recommended user pool are possibly far larger than the number recommended to the user A in each extraction, the user B in each user pool is sorted and the user B with the top rank is extracted when the user B is extracted from the three user pools, namely the personalized recommended pool, the one-way relation pool and the current hierarchy head user pool.
By tracking and recording the behavior data of the user A, for example, in a user B approved by the user A, the label is added to the user A if the user B has not only the label same as that of the user A but also a label which is not available to the user A;
for example, if the user a is exposed 10 times for the first time and "liked" for 5 times, and exposed 10 times for the second time and "liked" for 7 times, the floating score will float, the total score y will be correspondingly increased, and the hierarchy may also be increased.
The present invention is further explained and not limited by the embodiments, and those skilled in the art can make various modifications as necessary after reading the present specification, but all the embodiments are protected by the patent law within the scope of the claims.
Claims (8)
1. A user recommendation method in an internet matching social scene is characterized by comprising user A level division, user recommendation pool division, user B in a user recommendation pool extraction in a layered mode, user A preference behavior data collection, user A level division correction according to the behavior data of the user A, and the user B extracted from the user recommendation pool.
2. The user recommendation method in the internet matching social scene according to claim 1, wherein the user A is classified according to the grade of the user A during the hierarchy classification, and the grade of the user A comprises a basic grade b and a floating grade a; the basic score b is obtained by weighted calculation according to the basic information of the user A, and the floating score a is obtained by weighted calculation according to the times that the user A is preferred by other users/the times that the user A recommends to other users.
3. The user recommendation method under the internet matching social scene according to claim 1 or 2, wherein the user recommendation pool comprises a one-way relationship pool, a current level head user pool and a personalized recommendation pool, and the personalized recommendation pool is formed according to the similarity matching of a user A tag and a user B tag; the unidirectional relation pool is generated according to the preference behavior of the user B to the user A; and the current hierarchy head user pool is formed according to the hierarchy division of the user A, and the hierarchy of the user B in the current hierarchy head user pool is the same as that of the user A.
4. The method according to claim 3, wherein the priorities of the personalized recommendation pool, the one-way relationship pool and the current-level head user pool are sequentially reduced, and when the user B in the user recommendation pool is extracted hierarchically, the user B is sequentially extracted from the personalized recommendation pool, the one-way relationship pool and the current-level head user pool.
5. The method as claimed in claim 4, wherein the extraction ratio of the personalized recommendation pool is greater than that of the one-way relationship pool and the current-level head user pool, and the extraction ratio of the one-way relationship pool is greater than that of the current-level head user pool.
6. The method as claimed in claim 5, wherein when the user B is hierarchically extracted from the user recommendation pool, the extraction rate of the previous priority is insufficient to extract the complementary extraction rate from the next priority.
7. The method for recommending users in the internet matching social scene according to any of claims 4-6, wherein when a user B is extracted from the personalized recommendation pool, the one-way relationship pool and the current level head user pool, users B in the extracted personalized recommendation pool, the one-way relationship pool and the current level head user pool are ranked respectively, and users B with the highest ranking are extracted.
8. The method as claimed in claim 1, wherein the user B is a set of users recommended to the user A, and the set of users does not include the user A itself.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110030756.0A CN112559893A (en) | 2021-01-11 | 2021-01-11 | User recommendation method in internet matching social scene |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110030756.0A CN112559893A (en) | 2021-01-11 | 2021-01-11 | User recommendation method in internet matching social scene |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112559893A true CN112559893A (en) | 2021-03-26 |
Family
ID=75035425
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110030756.0A Pending CN112559893A (en) | 2021-01-11 | 2021-01-11 | User recommendation method in internet matching social scene |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112559893A (en) |
Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106708938A (en) * | 2016-11-18 | 2017-05-24 | 北京大米科技有限公司 | Method and device for assisting recommendation |
| CN106815216A (en) * | 2015-11-30 | 2017-06-09 | 北京云莱坞文化传媒有限公司 | A kind of story screening and the method and apparatus for precisely representing |
| CN106951515A (en) * | 2017-03-17 | 2017-07-14 | 上海衡修信息科技有限公司 | A kind of contact person's matching process and device based on social software |
| CN107526850A (en) * | 2017-10-12 | 2017-12-29 | 燕山大学 | Social networks friend recommendation method based on multiple personality feature mixed architecture |
| CN107767279A (en) * | 2017-10-12 | 2018-03-06 | 燕山大学 | A kind of average weighted personalized friend recommendation method based on LDA |
| CN108021645A (en) * | 2017-11-30 | 2018-05-11 | 华南理工大学 | It is a kind of based on the potential friend recommendation method for choosing friends preference and matrix decomposition |
| CN108427715A (en) * | 2018-01-30 | 2018-08-21 | 重庆邮电大学 | A kind of social networks friend recommendation method of fusion degree of belief |
| CN109190033A (en) * | 2018-08-23 | 2019-01-11 | 微梦创科网络科技（中国）有限公司 | A kind of user's friend recommendation method and system |
| CN110188123A (en) * | 2019-04-24 | 2019-08-30 | 上海任意门科技有限公司 | User matching method and equipment |
| CN110196951A (en) * | 2019-04-24 | 2019-09-03 | 上海任意门科技有限公司 | User matching method and equipment |
| CN110209704A (en) * | 2019-04-24 | 2019-09-06 | 上海任意门科技有限公司 | User matching method and equipment |
| CN110913249A (en) * | 2018-09-18 | 2020-03-24 | 深圳市茁壮网络股份有限公司 | Program recommendation method and system |
| CN111079009A (en) * | 2019-12-11 | 2020-04-28 | 中国地质大学（武汉） | User interest detection method and system for government map service |
-
2021
- 2021-01-11 CN CN202110030756.0A patent/CN112559893A/en active Pending
Patent Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106815216A (en) * | 2015-11-30 | 2017-06-09 | 北京云莱坞文化传媒有限公司 | A kind of story screening and the method and apparatus for precisely representing |
| CN106708938A (en) * | 2016-11-18 | 2017-05-24 | 北京大米科技有限公司 | Method and device for assisting recommendation |
| CN106951515A (en) * | 2017-03-17 | 2017-07-14 | 上海衡修信息科技有限公司 | A kind of contact person's matching process and device based on social software |
| CN107526850A (en) * | 2017-10-12 | 2017-12-29 | 燕山大学 | Social networks friend recommendation method based on multiple personality feature mixed architecture |
| CN107767279A (en) * | 2017-10-12 | 2018-03-06 | 燕山大学 | A kind of average weighted personalized friend recommendation method based on LDA |
| CN108021645A (en) * | 2017-11-30 | 2018-05-11 | 华南理工大学 | It is a kind of based on the potential friend recommendation method for choosing friends preference and matrix decomposition |
| CN108427715A (en) * | 2018-01-30 | 2018-08-21 | 重庆邮电大学 | A kind of social networks friend recommendation method of fusion degree of belief |
| CN109190033A (en) * | 2018-08-23 | 2019-01-11 | 微梦创科网络科技（中国）有限公司 | A kind of user's friend recommendation method and system |
| CN110913249A (en) * | 2018-09-18 | 2020-03-24 | 深圳市茁壮网络股份有限公司 | Program recommendation method and system |
| CN110188123A (en) * | 2019-04-24 | 2019-08-30 | 上海任意门科技有限公司 | User matching method and equipment |
| CN110196951A (en) * | 2019-04-24 | 2019-09-03 | 上海任意门科技有限公司 | User matching method and equipment |
| CN110209704A (en) * | 2019-04-24 | 2019-09-06 | 上海任意门科技有限公司 | User matching method and equipment |
| CN111079009A (en) * | 2019-12-11 | 2020-04-28 | 中国地质大学（武汉） | User interest detection method and system for government map service |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110555112B (en) | Interest point recommendation method based on user positive and negative preference learning | |
| CN110162700B (en) | Training method, device and equipment for information recommendation and model and storage medium | |
| Jiang et al. | Author topic model-based collaborative filtering for personalized POI recommendations | |
| Loken et al. | Brands and brand management: Contemporary research perspectives | |
| Elmer | Profiling machines: Mapping the personal information economy | |
| US9058611B2 (en) | System and method for advertising using image search and classification | |
| US20190242720A1 (en) | Systems and methods for constructing spatial activity zones | |
| Arabadzhyan et al. | Measuring destination image: a novel approach based on visual data mining. A methodological proposal and an application to European islands | |
| CN105976161A (en) | Time axis-based intelligent recommendation calendar and user-based presentation method | |
| US20120174006A1 (en) | System, method, apparatus and computer program for generating and modeling a scene | |
| CN106537390B (en) | Identify the presentation style of education video | |
| CN107220365A (en) | Accurate commending system and method based on collaborative filtering and correlation rule parallel processing | |
| CN105975581A (en) | Media information display method, client and server | |
| CN101510856A (en) | Method and apparatus for extracting member relation loop in SNS network | |
| JP2009076042A (en) | Learning user's activity preference from gps trace and known nearby venue | |
| CN108665083A (en) | A kind of method and system for advertisement recommendation for dynamic trajectory model of being drawn a portrait based on user | |
| Gao et al. | Mining human mobility in location-based social networks | |
| Prokopenko et al. | Digital-toolkit for sports tourism promoting | |
| CN112270579B (en) | Intelligent advertising system based on big data | |
| CN110070134A (en) | A kind of recommended method and device based on user interest perception | |
| CN107357845A (en) | A kind of tour interest commending system and recommendation method based on Spark | |
| CN105787069A (en) | Personalized music recommendation method | |
| CN112287241B (en) | Travel recommendation method and system | |
| CN112685596B (en) | Video recommendation method and device, terminal and storage medium | |
| CN110309363A (en) | A kind of instructional video segment method of commerce of knowledge based point |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20210326 |