WO2018032790A1 - Weighted k-nearest-neighbor scoring-based live broadcast room recommendation method and system - Google Patents

Weighted k-nearest-neighbor scoring-based live broadcast room recommendation method and system Download PDF

Info

Publication number
WO2018032790A1
WO2018032790A1 PCT/CN2017/080786 CN2017080786W WO2018032790A1 WO 2018032790 A1 WO2018032790 A1 WO 2018032790A1 CN 2017080786 W CN2017080786 W CN 2017080786W WO 2018032790 A1 WO2018032790 A1 WO 2018032790A1
Authority
WO
WIPO (PCT)
Prior art keywords
user
live broadcast
live
room
similarity
Prior art date
Application number
PCT/CN2017/080786
Other languages
French (fr)
Chinese (zh)
Inventor
龚灿
Original Assignee
武汉斗鱼网络科技有限公司
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 武汉斗鱼网络科技有限公司 filed Critical 武汉斗鱼网络科技有限公司
Publication of WO2018032790A1 publication Critical patent/WO2018032790A1/en

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
    • 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

Definitions

  • the invention relates to the field of promotion of a webcasting room, and particularly relates to a method and system for recommending a live broadcast based on a weighted k-nearest neighbor score.
  • the traditional live broadcast platform recommends the live broadcast room (ie, the link of the live content viewed) to the user.
  • the method is generally based on the k-nearest neighbor recommendation method, and the method is used to recommend the live broadcast room (ie, the link of the live content viewed) to the user A.
  • the specific process is: firstly, using the similarity calculation formula to select K users (such as B1, B2, ..., BK) similar to the interests of user A, and then merge and merge in the live broadcasts that K users like to watch, and merge
  • K users such as B1, B2, ..., BK
  • the recommended method based on the k-nearest neighbor is for the user to recommend the live broadcast room
  • the user A selects among the recommended lists of the live broadcasts of the K users, because different users of the K users are similar to the user A.
  • the degree (viewing grade) is different, so it is difficult for user A to quickly find the live room recommended by the user with the closest or the same degree of similarity in the recommendation list; because the similarity is higher, the recommended live room is The more accurate, therefore, the existing k-nearest neighbor-based recommendation method is difficult to quickly recommend a high-level live broadcast room for the user, that is, the accuracy of the recommendation is not enough, and the recommendation effect needs to be improved.
  • the technical problem solved by the present invention is to improve the accuracy of the live broadcast recommended to the user, and to quickly recommend the live broadcast room that meets the user's interests and individual needs.
  • the method for recommending live broadcasts based on the weighted k-nearest neighbor score provided by the present invention includes the following steps:
  • S1 Obtain a live broadcast view vector of each user in the live broadcast platform, and the live broadcast view vector includes: a user views the set of views of each live broadcast room;
  • R mn represents the similarity between the user m and the user n
  • represents the absolute value of R mn
  • N i (m) represents the set of neighboring users of the user m
  • S ni represents the user n to the live room i Times watched
  • the recommendation system includes a live view vector acquisition module, a similarity calculation module, a neighboring user selection module, an interest degree calculation module, and a recommended live broadcast selection module;
  • the live view vector acquisition module is configured to: obtain a live view vector of each user in the live broadcast platform, and the live view vector includes: the user views the set of views of each live broadcast;
  • the similarity calculation module is configured to: calculate a similarity degree of each of the two users according to the live broadcast viewing vector acquired by the live view vector acquisition module;
  • the neighboring user selection module is configured to: select, according to the similarity descending order of each user, the former N neighboring users as neighboring users, N neighbors >5;
  • the interest degree calculation module is configured to: calculate, according to the similarity and the neighboring users of each user, the degree of interest of each user m for each viewed live room i
  • the calculation formula is:
  • R mn represents the similarity between the user m and the user n
  • represents the absolute value of R mn
  • N i (m) represents the set of neighboring users of the user m
  • S ni represents the user n to the live room i Times watched
  • the recommended live room selection module is used to: for each user to sort according to the degree of interest in descending order, select the top N push live broadcast room as the recommended live broadcast room, N push >5.
  • the present invention is based on a weighted k-nearest neighbor score, and calculates a neighboring user of a recommended user (ie, a user who needs to recommend a live broadcast room) by a self-developed formula, and calculates a recommended user's view of each live broadcast room according to the similarity of the neighboring users.
  • Interest level Compared with the prior art, it is difficult to quickly recommend an accurate live broadcast to the user, and the present invention evaluates and selects the recommended live broadcast according to the interest calculated by the similarity, and the recommended live broadcast is more in line with the user's interests. And the personalized needs, the recommended accuracy is higher, the recommended effect is better.
  • FIG. 1 is a flowchart of a method for recommending a live broadcast based on a weighted k-nearest neighbor score according to an embodiment of the present invention.
  • a method for recommending a live broadcast based on a weighted k-neighbor score in an embodiment of the present invention includes the following steps:
  • S1 Obtain a live broadcast view vector of each user in the live broadcast platform, and the live broadcast view vector includes: the user views the set of views of each live broadcast room.
  • S101 Obtain historical viewing information within a specified period of each user (for example, 30 days).
  • the main fields of the historical viewing information include a UID (user unique identifier), and a ROOM_ID (a unique identifier of the viewed live broadcast room) associated with the UID.
  • S2 Calculate the similarity R of each of the two users according to the viewing vector X of the live broadcast platform among the users of the live broadcast platform. For example, if there are 56 users in the live broadcast platform, the calculation is performed for 56 users respectively; Other users than users, the choice is not heavy The other users of the complex calculation calculate the similarity between the other user and the calculated user.
  • X u and X v represent the live broadcast viewing vectors of two users u and v, respectively
  • T represents matrix transposition
  • represent X u and X v respectively . mold.
  • R mn represents the similarity between the user m and the user n, that is, the similarity is used as the weight
  • represents the absolute value of R mn
  • N i (m) represents the set of neighboring users of the user m (ie, the user n In the set of neighboring users of the user m
  • S ni represents the number of times the user n views the live room i.
  • the sum of weights with similarity as weight is not necessarily 1, it is divided by The purpose is to standardize the newly generated interest.
  • N push live broadcast room As the recommended live broadcast room, and N push >5; for example, the number of recommended live broadcast rooms in this embodiment is 10, that is, for each user, select the top 10 live broadcast rooms with high interest level as the recommended live broadcast room.
  • S6 Form a live broadcast recommendation list for each user's recommended live broadcast room and display it to the corresponding user.
  • the live broadcast recommendation system based on the weighted k-nearest neighbor score which comprises the above-mentioned method, includes a live view vector acquisition module, a similarity calculation module, a neighboring user selection module, an interest degree calculation module, a recommended live broadcast selection module, and a live broadcast room.
  • Recommended list acquisition module includes a live view vector acquisition module, a similarity calculation module, a neighboring user selection module, an interest degree calculation module, a recommended live broadcast selection module, and a live broadcast room.
  • the live view vector acquisition module is configured to: obtain a live broadcast view vector of each user in the live broadcast platform (the user views the set of views of each live broadcast room); the specific workflow is: obtaining the history of each user within a specified period Viewing information (UID, and the unique identifier ROOM_ID between the live broadcasts associated with the UID), clearing the invalid historical viewing information (clearing the invalid UID, clearing the data with empty ROOM_ID), and obtaining valid historical viewing information; according to the valid ROOM_ID, The number of views of the live room corresponding to each ROOM_ID viewed by the user is counted, and the set of all views forms the live view vector of the user.
  • Viewing information UID
  • ROOM_ID the unique identifier
  • the similarity calculation module is configured to calculate the similarity degree of each of the two users according to the live broadcast viewing vector acquired by the live view vector acquisition module; the similarity R is calculated as:
  • X u and X v represent the live broadcast viewing vectors of two users u and v, respectively
  • T represents matrix transposition
  • represent X u and X v respectively . mold.
  • the neighboring user selection module is configured to: select, according to the similarity descending order of each user, the first N neighboring users as neighboring users, N neighbors >5;
  • the interest degree calculation module is configured to: calculate, according to the similarity and the neighboring users of each user, the degree of interest of each user m for each viewed live room i
  • the calculation formula is:
  • R mn represents the similarity between the user m and the user n
  • represents the absolute value of R mn
  • N i (m) represents the set of neighboring users of the user m
  • S ni represents the user n to the live room i Times watched
  • the recommended live room selection module is used to: for each user to sort according to the degree of interest in descending order, select the top N push live broadcast room as the recommended live broadcast room, N push >5.
  • the live broadcast recommendation list obtaining module is configured to form a live broadcast recommendation list for each recommended live broadcast of each user.

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)
  • Two-Way Televisions, Distribution Of Moving Picture Or The Like (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

A weighted k-nearest-neighbor scoring-based live broadcast room recommendation method and system, relating to the field of online live broadcast room popularization. The steps of the method comprise: obtaining a live broadcast room watching vector of each user in a live broadcast platform (S1); calculating a similarity between every two users according to the live broadcast room watching vectors (S2); selecting the appointed number of neighbor users for each user according to a similarity descending arrangement method (S3); calculating, according to the similarities and the neighbor users of each user, an interest degree of each user in each watched live broadcast room (S4); and selecting the appointed number of live broadcast rooms to be recommended for each user according to an interest degree descending arrangement method (S5). The method and system evaluates and selects live broadcast rooms to be recommended according to an interest degree calculated based on similarities, such that the recommended live broadcast rooms can better meet user's interest and hobbies and personalized demands, the recommendation accuracy becomes higher, and the recommendation effect gets better.

Description

基于加权k近邻评分的直播间推荐方法及系统Method and system for recommending live broadcast based on weighted k-nearest neighbor score 技术领域Technical field
本发明涉及网络直播间的推广领域，具体涉及一种基于加权k近邻评分的直播间推荐方法及系统。The invention relates to the field of promotion of a webcasting room, and particularly relates to a method and system for recommending a live broadcast based on a weighted k-nearest neighbor score.
背景技术Background technique
随着智能终端的多屏化发展，中国社交视频的直播平台的活跃用户正在不断发展壮大中，人们对“即时观看喜爱的直播内容并与主播互动”的需求越来越高。因此，如何发掘用户兴趣点、给用户精准推荐直播间来提高用户粘性、促进用户的付费转化，将是直播行业很长一段时间将要面临的一道难题。With the multi-screen development of smart terminals, the active users of the live broadcast platform of social video in China are constantly growing and growing, and people are increasingly demanding “instantly watching favorite live content and interacting with the anchor”. Therefore, how to explore user interest points, accurately recommend live broadcasts to users to improve user stickiness, and promote users' payment conversion will be a difficult problem that the live broadcast industry will face for a long time.
目前，传统的直播平台为用户推荐直播间(即观看的直播内容的链接)的方法一般为基于k近邻的推荐方法，采用该方法为用户A推荐直播间(即观看的直播内容的链接)的具体流程为：首先利用相似度计算公式选出与用户A的兴趣爱好相似K个用户(例如B1,B2,…,BK)，然后在K个用户最喜欢观看的直播间并进行合并，将合并的直播间以推荐列表的形式呈现给用户A，供用户A选择观看。At present, the traditional live broadcast platform recommends the live broadcast room (ie, the link of the live content viewed) to the user. The method is generally based on the k-nearest neighbor recommendation method, and the method is used to recommend the live broadcast room (ie, the link of the live content viewed) to the user A. The specific process is: firstly, using the similarity calculation formula to select K users (such as B1, B2, ..., BK) similar to the interests of user A, and then merge and merge in the live broadcasts that K users like to watch, and merge The live room is presented to the user A in the form of a recommendation list for the user A to select to view.
但是，上述基于k近邻的推荐方法为用户推荐直播间时，存在以下缺陷：用户A在K个用户喜爱的直播间形成推荐列表中进行选择时，因为K个用户中不同用户与用户A的相似程度(观看品位)不同，所以用户A难以在推荐列表中，快速找到与自己相似程度最相近或相同的用户推荐的直播间；由于相似程度越高，推荐的直播间就 越准确，因此，现有的基于k近邻的推荐方法难以快速为用户推荐相似程度高的直播间，即推荐的准确度不够，推荐效果有待提高。However, when the recommended method based on the k-nearest neighbor is for the user to recommend the live broadcast room, there is a defect that the user A selects among the recommended lists of the live broadcasts of the K users, because different users of the K users are similar to the user A. The degree (viewing grade) is different, so it is difficult for user A to quickly find the live room recommended by the user with the closest or the same degree of similarity in the recommendation list; because the similarity is higher, the recommended live room is The more accurate, therefore, the existing k-nearest neighbor-based recommendation method is difficult to quickly recommend a high-level live broadcast room for the user, that is, the accuracy of the recommendation is not enough, and the recommendation effect needs to be improved.
发明内容Summary of the invention
针对现有技术中存在的缺陷，本发明解决的技术问题为：提高推荐给用户的直播间的准确度，进而为用户快速的推荐符合用户兴趣爱好和个性化需求的直播间。The technical problem solved by the present invention is to improve the accuracy of the live broadcast recommended to the user, and to quickly recommend the live broadcast room that meets the user's interests and individual needs.
为达到以上目的，本发明提供的基于加权k近邻评分的直播间推荐方法，包括以下步骤：To achieve the above objective, the method for recommending live broadcasts based on the weighted k-nearest neighbor score provided by the present invention includes the following steps:
S1：获取直播平台中每个用户的直播间观看向量，直播间观看向量包括：用户观看每个直播间的观看次数集合；S1: Obtain a live broadcast view vector of each user in the live broadcast platform, and the live broadcast view vector includes: a user views the set of views of each live broadcast room;
S2：根据直播间观看向量，计算每2个用户的相似度；S2: calculating the similarity of each of the two users according to the viewing vector between the live broadcasts;
S3：为所述每个用户按照相似度降序排列的方法，选取前N邻个用户作为邻近用户，N邻＞5；S3: For each user in descending order of similarity, select the N neighboring users as neighboring users, and N neighbors >5;
S4：根据相似度和所述每个用户的邻近用户，计算每个用户m对每个观看过的直播间i的兴趣度
Figure PCTCN2017080786-appb-000001
计算公式为：S4: Calculate the degree of interest of each user m for each viewed live room i according to the similarity and the neighboring users of each user.
Figure PCTCN2017080786-appb-000001
The calculation formula is:
Figure PCTCN2017080786-appb-000002
Figure PCTCN2017080786-appb-000002
上述公式中，Rmn代表用户m和用户n的相似度，|Rmn|代表Rmn的绝对值；Ni(m)代表用户m的邻近用户集合，Sni代表用户n对直播间i的观看次数；In the above formula, R mn represents the similarity between the user m and the user n, |R mn | represents the absolute value of R mn ; N i (m) represents the set of neighboring users of the user m, and S ni represents the user n to the live room i Times watched;
S5：为所述每个用户按照兴趣度降序排列的方法，选取前N推个直播间作为推荐直播间，N推＞5。S5: For each user to sort according to the degree of interest in descending order, select the first N push live broadcast room as the recommended live broadcast room, and N push >5.
本发明提供的实现上述方法的基于加权k近邻评分的直播间推 荐系统，包括直播间观看向量获取模块、相似度计算模块、邻近用户选取模块、兴趣度计算模块和推荐直播间选取模块；The live broadcast push based on the weighted k-nearest neighbor score for implementing the above method provided by the present invention The recommendation system includes a live view vector acquisition module, a similarity calculation module, a neighboring user selection module, an interest degree calculation module, and a recommended live broadcast selection module;
直播间观看向量获取模块用于：获取直播平台中每个用户的直播间观看向量，直播间观看向量包括：用户观看每个直播间的观看次数集合；The live view vector acquisition module is configured to: obtain a live view vector of each user in the live broadcast platform, and the live view vector includes: the user views the set of views of each live broadcast;
相似度计算模块用于：根据直播间观看向量获取模块获取的直播间观看向量，计算每2个用户的相似度；The similarity calculation module is configured to: calculate a similarity degree of each of the two users according to the live broadcast viewing vector acquired by the live view vector acquisition module;
邻近用户选取模块用于：为所述每个用户按照相似度降序排列的方法，选取前N邻个用户作为邻近用户，N邻＞5；The neighboring user selection module is configured to: select, according to the similarity descending order of each user, the former N neighboring users as neighboring users, N neighbors >5;
兴趣度计算模块用于：根据相似度和所述每个用户的邻近用户，计算每个用户m对每个观看过的直播间i的兴趣度
Figure PCTCN2017080786-appb-000003
计算公式为：The interest degree calculation module is configured to: calculate, according to the similarity and the neighboring users of each user, the degree of interest of each user m for each viewed live room i
Figure PCTCN2017080786-appb-000003
The calculation formula is:
Figure PCTCN2017080786-appb-000004
Figure PCTCN2017080786-appb-000004
上述公式中，Rmn代表用户m和用户n的相似度，|Rmn|代表Rmn的绝对值；Ni(m)代表用户m的邻近用户集合，Sni代表用户n对直播间i的观看次数；In the above formula, R mn represents the similarity between the user m and the user n, |R mn | represents the absolute value of R mn ; N i (m) represents the set of neighboring users of the user m, and S ni represents the user n to the live room i Times watched;
推荐直播间选取模块用于：为所述每个用户按照兴趣度降序排列的方法，选取前N推个直播间作为推荐直播间，N推＞5。The recommended live room selection module is used to: for each user to sort according to the degree of interest in descending order, select the top N push live broadcast room as the recommended live broadcast room, N push >5.
与现有技术相比，本发明的优点在于：The advantages of the present invention over the prior art are:
本发明基于加权k近邻评分，通过自主研发的公式计算出被推荐用户(即需要推荐直播间的用户)的邻近用户，根据邻近用户的相似度计算被推荐用户对观看过的每个直播间的兴趣度。与现有技术中难以快速为用户推荐准确的直播间相比，本发明根据由相似度计算的兴趣度评估和选取推荐直播间，推荐的直播间更加符合用户的兴趣爱好 和个性化需求，推荐的准确度较高，推荐效果较好。The present invention is based on a weighted k-nearest neighbor score, and calculates a neighboring user of a recommended user (ie, a user who needs to recommend a live broadcast room) by a self-developed formula, and calculates a recommended user's view of each live broadcast room according to the similarity of the neighboring users. Interest level. Compared with the prior art, it is difficult to quickly recommend an accurate live broadcast to the user, and the present invention evaluates and selects the recommended live broadcast according to the interest calculated by the similarity, and the recommended live broadcast is more in line with the user's interests. And the personalized needs, the recommended accuracy is higher, the recommended effect is better.
附图说明DRAWINGS
图1为本发明实施例中基于加权k近邻评分的直播间推荐方法的流程图。FIG. 1 is a flowchart of a method for recommending a live broadcast based on a weighted k-nearest neighbor score according to an embodiment of the present invention.
具体实施方式detailed description
以下结合附图及实施例对本发明作进一步详细说明。The present invention will be further described in detail below with reference to the accompanying drawings and embodiments.
参见图1所示，本发明实施例中的基于加权k近邻评分的直播间推荐方法，包括以下步骤：Referring to FIG. 1 , a method for recommending a live broadcast based on a weighted k-neighbor score in an embodiment of the present invention includes the following steps:
S1：获取直播平台中每个用户的直播间观看向量，直播间观看向量包括：用户观看每个直播间的观看次数集合。S1: Obtain a live broadcast view vector of each user in the live broadcast platform, and the live broadcast view vector includes: the user views the set of views of each live broadcast room.
S1的具体流程为：The specific process of S1 is:
S101：获取每个用户指定期限内(例如30天)的历史观看信息，历史观看信息的主要字段包括UID(用户唯一标识)、以及与UID关联的ROOM_ID(观看过的直播间的唯一标识)。S101: Obtain historical viewing information within a specified period of each user (for example, 30 days). The main fields of the historical viewing information include a UID (user unique identifier), and a ROOM_ID (a unique identifier of the viewed live broadcast room) associated with the UID.
S102：在历史观看信息中，清除无效的历史观看信息(即清除无效UID和ROOM_ID为空的数据)后，得到有效的历史观看信息。根据有效的历史观看信息(ROOM_ID)，统计用户观看的每个ROOM_ID对应的直播间的观看次数，所有观看次数的集合形成该用户的直播间观看向量X。例如X为：a1,a2,a3,a4,…，aN，则代表用户a分别对N个直播间的观看次数，a1表示用户a对直播间1的观看次数，以此类推。S102: In the historical viewing information, after clearing the invalid historical viewing information (ie, clearing the invalid UID and the ROOM_ID is empty data), valid historical viewing information is obtained. According to the valid historical viewing information (ROOM_ID), the number of views of the live room corresponding to each ROOM_ID viewed by the user is counted, and the set of all the viewing times forms the live view vector X of the user. For example, X is: a1, a2, a3, a4, ..., aN, which represents the number of views of the user a to the N live broadcasts, a1 represents the number of views of the user a to the live room 1, and so on.
S2：在直播平台的所用用户中，根据直播间观看向量X，计算每2个用户的相似度R，例如直播平台中有56个用户，则分别针对56个用户进行计算；计算时遍历除计算用户之外的其他用户，选择未重 复计算的其他用户，计算该其他用户与计算用户的相似度。S2: Calculate the similarity R of each of the two users according to the viewing vector X of the live broadcast platform among the users of the live broadcast platform. For example, if there are 56 users in the live broadcast platform, the calculation is performed for 56 users respectively; Other users than users, the choice is not heavy The other users of the complex calculation calculate the similarity between the other user and the calculated user.
S2中相似度R的计算公式为：The formula for calculating the similarity R in S2 is:
Figure PCTCN2017080786-appb-000005
Figure PCTCN2017080786-appb-000005
上述公式中，Xu和Xv分别代表2个用户u和v的直播间观看向量，T代表矩阵转置，||Xu||和||Xv||分别代表Xu和Xv的模。In the above formula, X u and X v represent the live broadcast viewing vectors of two users u and v, respectively, T represents matrix transposition, and ||X u || and ||X v || represent X u and X v respectively . mold.
Figure PCTCN2017080786-appb-000006
其实就是计算Xu和Xv的余弦夹角，从几何角度上，向量夹角越小，表示向量相似度越大。
Figure PCTCN2017080786-appb-000006
In fact, it is to calculate the cosine angle of X u and X v . From the geometric point of view, the smaller the angle of the vector, the greater the similarity of the vector.
S3：在直播平台的所用用户中，为所述每个用户按照相似度降序排列的方法，选取前N邻个用户作为邻近用户，N邻＞5；例如本实施例中邻近用户的数量为12个，即为每个用户选取前12个相似度大的用户作为邻近用户。S3: In the broadcast platform by users, each user of the method according to the degree of similarity in descending order, selecting a first N o users near the user, N o> 5; example embodiment of the present embodiment the number of the user 12 adjacent the For each user, the first 12 users with similar similarities are selected as neighboring users.
S4：在直播平台的所用用户中，根据相似度R，计算每个用户m对每个观看过的直播间i的兴趣度
Figure PCTCN2017080786-appb-000007
计算公式为：S4: Calculating the interest of each user m on each viewed live room i according to the similarity R among the users of the live platform
Figure PCTCN2017080786-appb-000007
The calculation formula is:
Figure PCTCN2017080786-appb-000008
Figure PCTCN2017080786-appb-000008
上述公式中，Rmn代表用户m和用户n的相似度，即以相似度作为权重，|Rmn|代表Rmn的绝对值；Ni(m)代表用户m的邻近用户集合(即用户n在用户m的邻近用户集合中选取)，Sni代表用户n对直播间i的观看次数。考虑到以相似度作为权重的权重总和不一定为1，所以这里除以
Figure PCTCN2017080786-appb-000009
的目的是对新生成的兴趣度作标准化处理。 In the above formula, R mn represents the similarity between the user m and the user n, that is, the similarity is used as the weight, |R mn | represents the absolute value of R mn ; N i (m) represents the set of neighboring users of the user m (ie, the user n In the set of neighboring users of the user m, S ni represents the number of times the user n views the live room i. Considering that the sum of weights with similarity as weight is not necessarily 1, it is divided by
Figure PCTCN2017080786-appb-000009
The purpose is to standardize the newly generated interest.
S5：在直播平台的所用用户中，为每个用户按照兴趣度降序排列的方法，选取前N推个直播间作为推荐直播间，N推＞5；例如本实施例中推荐直播间的数量为10个，即为每个用户选取前10个兴趣度大的直播间作为推荐直播间。S5: In the user of the live broadcast platform, for each user, according to the method of descending order of interest, select the first N push live broadcast room as the recommended live broadcast room, and N push >5; for example, the number of recommended live broadcast rooms in this embodiment is 10, that is, for each user, select the top 10 live broadcast rooms with high interest level as the recommended live broadcast room.
S6：将每个用户的推荐直播间形成直播间推荐列表、并展示给相应的用户。S6: Form a live broadcast recommendation list for each user's recommended live broadcast room and display it to the corresponding user.
本发明提供的实现上述方法的基于加权k近邻评分的直播间推荐系统，包括直播间观看向量获取模块、相似度计算模块、邻近用户选取模块、兴趣度计算模块、推荐直播间选取模块和直播间推荐列表获取模块。The live broadcast recommendation system based on the weighted k-nearest neighbor score, which comprises the above-mentioned method, includes a live view vector acquisition module, a similarity calculation module, a neighboring user selection module, an interest degree calculation module, a recommended live broadcast selection module, and a live broadcast room. Recommended list acquisition module.
直播间观看向量获取模块用于：获取直播平台中每个用户的直播间观看向量(用户观看每个直播间的观看次数集合)；具体工作流程为：获取所述每个用户指定期限内的历史观看信息(UID、以及与UID关联的直播间唯一标识ROOM_ID)，清除无效的历史观看信息后(清除无效的UID，清楚ROOM_ID为空的数据)，得到有效的历史观看信息；根据有效的ROOM_ID，统计用户观看的每个ROOM_ID对应的直播间的观看次数，所有观看次数的集合形成该用户的直播间观看向量。The live view vector acquisition module is configured to: obtain a live broadcast view vector of each user in the live broadcast platform (the user views the set of views of each live broadcast room); the specific workflow is: obtaining the history of each user within a specified period Viewing information (UID, and the unique identifier ROOM_ID between the live broadcasts associated with the UID), clearing the invalid historical viewing information (clearing the invalid UID, clearing the data with empty ROOM_ID), and obtaining valid historical viewing information; according to the valid ROOM_ID, The number of views of the live room corresponding to each ROOM_ID viewed by the user is counted, and the set of all views forms the live view vector of the user.
相似度计算模块用于：根据直播间观看向量获取模块获取的直播间观看向量，计算每2个用户的相似度；相似度R的计算公式为：The similarity calculation module is configured to calculate the similarity degree of each of the two users according to the live broadcast viewing vector acquired by the live view vector acquisition module; the similarity R is calculated as:
Figure PCTCN2017080786-appb-000010
Figure PCTCN2017080786-appb-000010
上述公式中，Xu和Xv分别代表2个用户u和v的直播间观看向量，T代表矩阵转置，||Xu||和||Xv||分别代表Xu和Xv的模。In the above formula, X u and X v represent the live broadcast viewing vectors of two users u and v, respectively, T represents matrix transposition, and ||X u || and ||X v || represent X u and X v respectively . mold.
邻近用户选取模块用于：为所述每个用户按照相似度降序排列的 方法，选取前N邻个用户作为邻近用户，N邻＞5；The neighboring user selection module is configured to: select, according to the similarity descending order of each user, the first N neighboring users as neighboring users, N neighbors >5;
兴趣度计算模块用于：根据相似度和所述每个用户的邻近用户，计算每个用户m对每个观看过的直播间i的兴趣度
Figure PCTCN2017080786-appb-000011
计算公式为：The interest degree calculation module is configured to: calculate, according to the similarity and the neighboring users of each user, the degree of interest of each user m for each viewed live room i
Figure PCTCN2017080786-appb-000011
The calculation formula is:
Figure PCTCN2017080786-appb-000012
Figure PCTCN2017080786-appb-000012
上述公式中，Rmn代表用户m和用户n的相似度，|Rmn|代表Rmn的绝对值；Ni(m)代表用户m的邻近用户集合，Sni代表用户n对直播间i的观看次数；In the above formula, R mn represents the similarity between the user m and the user n, |R mn | represents the absolute value of R mn ; N i (m) represents the set of neighboring users of the user m, and S ni represents the user n to the live room i Times watched;
推荐直播间选取模块用于：为所述每个用户按照兴趣度降序排列的方法，选取前N推个直播间作为推荐直播间，N推＞5。The recommended live room selection module is used to: for each user to sort according to the degree of interest in descending order, select the top N push live broadcast room as the recommended live broadcast room, N push >5.
直播间推荐列表获取模块用于：将所述每个用户的推荐直播间形成直播间推荐列表。The live broadcast recommendation list obtaining module is configured to form a live broadcast recommendation list for each recommended live broadcast of each user.
本发明不局限于上述实施方式，对于本技术领域的普通技术人员来说，在不脱离本发明原理的前提下，还可以做出若干改进和润饰，这些改进和润饰也视为本发明的保护范围之内。本说明书中未作详细描述的内容属于本领域专业技术人员公知的现有技术。 The present invention is not limited to the above embodiments, and those skilled in the art can also make several improvements and retouchings without departing from the principles of the present invention. These improvements and retouchings are also considered as protection of the present invention. Within the scope. The contents not described in detail in the present specification belong to the prior art well known to those skilled in the art.

Claims (10)

  1. 一种基于加权k近邻评分的直播间推荐方法，其特征在于，该方法包括以下步骤：A method for recommending a live broadcast based on a weighted k-nearest neighbor score, characterized in that the method comprises the following steps:
    S1：获取直播平台中每个用户的直播间观看向量，直播间观看向量包括：用户观看每个直播间的观看次数集合；S1: Obtain a live broadcast view vector of each user in the live broadcast platform, and the live broadcast view vector includes: a user views the set of views of each live broadcast room;
    S2：根据直播间观看向量，计算每2个用户的相似度；S2: calculating the similarity of each of the two users according to the viewing vector between the live broadcasts;
    S3：为所述每个用户按照相似度降序排列的方法，选取前N邻个用户作为邻近用户，N邻＞5；S3: For each user in descending order of similarity, select the N neighboring users as neighboring users, and N neighbors >5;
    S4：根据相似度和所述每个用户的邻近用户，计算每个用户m对每个观看过的直播间i的兴趣度
    Figure PCTCN2017080786-appb-100001
    计算公式为：    S4: Calculate the degree of interest of each user m for each viewed live room i according to the similarity and the neighboring users of each user.
    Figure PCTCN2017080786-appb-100001
    The calculation formula is:
    Figure PCTCN2017080786-appb-100002
    Figure PCTCN2017080786-appb-100002
    上述公式中，Rmn代表用户m和用户n的相似度，|Rmn|代表Rmn的绝对值；Ni(m)代表用户m的邻近用户集合，Sni代表用户n对直播间i的观看次数；In the above formula, R mn represents the similarity between the user m and the user n, |R mn | represents the absolute value of R mn ; N i (m) represents the set of neighboring users of the user m, and S ni represents the user n to the live room i Times watched;
    S5：为所述每个用户按照兴趣度降序排列的方法，选取前N推个直播间作为推荐直播间，N推＞5。S5: For each user to sort according to the degree of interest in descending order, select the first N push live broadcast room as the recommended live broadcast room, and N push >5.
  2. 如权利要求1所述的基于加权k近邻评分的直播间推荐方法，其特征在于，S1的具体流程为：获取所述每个用户指定期限内的历史观看信息，清除无效的历史观看信息后，得到有效的历史观看信息；根据有效的历史观看信息，统计每个用户的直播间观看向量。The method for recommending a live broadcast based on the weighted k-nearest neighbor score according to claim 1, wherein the specific process of S1 is: acquiring historical viewing information within a specified period of each user, and clearing invalid historical viewing information, Obtain effective historical viewing information; count the viewing vector of each user's live broadcast based on valid historical viewing information.
  3. 如权利要求2所述的基于加权k近邻评分的直播间推荐方法，其特征在于：所述历史观看信息包括UID、以及与UID关联的直播 间唯一标识ROOM_ID；在此基础上：The method for recommending live broadcasts based on weighted k-nearest neighbor score according to claim 2, wherein the historical viewing information comprises a UID and a live broadcast associated with the UID. Uniquely identifies ROOM_ID; on this basis:
    所述清除无效的历史观看信息的流程为：清除无效的UID，清楚ROOM_ID为空的数据；The process of clearing the invalid historical viewing information is: clearing the invalid UID, and clearing the data that the ROOM_ID is empty;
    所述根据有效的历史观看信息，统计每个用户的直播间观看向量的流程为：根据有效的ROOM_ID，统计用户观看的每个ROOM_ID对应的直播间的观看次数，所有观看次数的集合形成该用户的直播间观看向量。According to the effective historical viewing information, the process of counting the viewing vector of the live broadcast of each user is: according to the valid ROOM_ID, the number of views of the live broadcast corresponding to each ROOM_ID viewed by the user is counted, and the set of all the viewing times forms the user. Live room viewing vector.
  4. 如权利要求1所述的基于加权k近邻评分的直播间推荐方法，其特征在于，S2中相似度R的计算公式为：The live-between recommendation method based on the weighted k-nearest neighbor score according to claim 1, wherein the calculation formula of the similarity R in S2 is:
    Figure PCTCN2017080786-appb-100003
    Figure PCTCN2017080786-appb-100003
    上述公式中，Xu和Xv分别代表2个用户u和v的直播间观看向量，T代表矩阵转置，||Xu||和||Xv||分别代表Xu和Xv的模。In the above formula, X u and X v represent the live broadcast viewing vectors of two users u and v, respectively, T represents matrix transposition, and ||X u || and ||X v || represent X u and X v respectively . mold.
  5. 如权利要求1至4任一项所述的基于加权k近邻评分的直播间推荐方法，其特征在于，S5之后还包括以下步骤：S6：将所述每个用户的推荐直播间形成直播间推荐列表。The method for recommending a live broadcast based on the weighted k-nearest neighbor score according to any one of claims 1 to 4, further comprising the following steps: S6: forming a live broadcast recommendation between the recommended live broadcasts of each user. List.
  6. 一种实现权利要求1至5任一项所述方法的基于加权k近邻评分的直播间推荐系统，其特征在于：该系统包括直播间观看向量获取模块、相似度计算模块、邻近用户选取模块、兴趣度计算模块和推荐直播间选取模块；A live-between recommendation system based on weighted k-nearest neighbor scores for implementing the method according to any one of claims 1 to 5, characterized in that the system comprises a live view vector acquisition module, a similarity calculation module, a neighboring user selection module, The interest degree calculation module and the recommended live broadcast selection module;
    直播间观看向量获取模块用于：获取直播平台中每个用户的直播间观看向量，直播间观看向量包括：用户观看每个直播间的观看次数集合；The live view vector acquisition module is configured to: obtain a live view vector of each user in the live broadcast platform, and the live view vector includes: the user views the set of views of each live broadcast;
    相似度计算模块用于：根据直播间观看向量获取模块获取的直播间观看向量，计算每2个用户的相似度；The similarity calculation module is configured to: calculate a similarity degree of each of the two users according to the live broadcast viewing vector acquired by the live view vector acquisition module;
    邻近用户选取模块用于：为所述每个用户按照相似度降序排列的 方法，选取前N邻个用户作为邻近用户，N邻＞5；The neighboring user selection module is configured to: select, according to the similarity descending order of each user, the first N neighboring users as neighboring users, N neighbors >5;
    兴趣度计算模块用于：根据相似度和所述每个用户的邻近用户，计算每个用户m对每个观看过的直播间i的兴趣度
    Figure PCTCN2017080786-appb-100004
    计算公式为：    The interest degree calculation module is configured to: calculate, according to the similarity and the neighboring users of each user, the degree of interest of each user m for each viewed live room i
    Figure PCTCN2017080786-appb-100004
    The calculation formula is:
    Figure PCTCN2017080786-appb-100005
    Figure PCTCN2017080786-appb-100005
    上述公式中，Rmn代表用户m和用户n的相似度，|Rmn|代表Rmn的绝对值；Ni(m)代表用户m的邻近用户集合，Sni代表用户n对直播间i的观看次数；In the above formula, R mn represents the similarity between the user m and the user n, |R mn | represents the absolute value of R mn ; N i (m) represents the set of neighboring users of the user m, and S ni represents the user n to the live room i Times watched;
    推荐直播间选取模块用于：为所述每个用户按照兴趣度降序排列的方法，选取前N推个直播间作为推荐直播间，N推＞5。The recommended live room selection module is used to: for each user to sort according to the degree of interest in descending order, select the top N push live broadcast room as the recommended live broadcast room, N push >5.
  7. 如权利要求6所述的基于加权k近邻评分的直播间推荐系统，其特征在于，所述直播间观看向量获取模块的具体工作流程为：获取所述每个用户指定期限内的历史观看信息，清除无效的历史观看信息后，得到有效的历史观看信息；根据有效的历史观看信息，统计每个用户的直播间观看向量。The live broadcast recommendation system based on the weighted k-nearest neighbor score according to claim 6, wherein the specific workflow of the live broadcast view vector acquisition module is: acquiring historical viewing information within a specified period of each user. After clearing the invalid historical viewing information, effective historical viewing information is obtained; and according to the effective historical viewing information, the live broadcast viewing vector of each user is counted.
  8. 如权利要求7所述的基于加权k近邻评分的直播间推荐系统，其特征在于：所述历史观看信息包括UID、以及与UID关联的直播间唯一标识ROOM_ID；在此基础上：The live-between recommendation system based on the weighted k-nearest neighbor score according to claim 7, wherein the historical viewing information comprises a UID and a unique identifier ROOM_ID between the live broadcasts associated with the UID;
    所述直播间观看向量获取模块清除无效的历史观看信息的工作流程为：清除无效的UID，清楚ROOM_ID为空的数据；The workflow of the live view vector acquisition module to clear the invalid historical viewing information is: clearing the invalid UID, and clearing the data that the ROOM_ID is empty;
    所述直播间观看向量获取模块根据有效的历史观看信息，统计每个用户的直播间观看向量的工作流程为：根据有效的ROOM_ID，统计用户观看的每个ROOM_ID对应的直播间的观看次数，所有观看次数的集合形成该用户的直播间观看向量。 The live view vector acquisition module calculates the workflow of the live view vector of each user according to the valid historical view information: according to the valid ROOM_ID, the number of views of the live broadcast corresponding to each ROOM_ID viewed by the user is counted, all The set of views forms the live view vector of the user.
  9. 如权利要求6所述的基于加权k近邻评分的直播间推荐系统，其特征在于，所述相似度计算模块中相似度R的计算公式为：The live-between recommendation system based on the weighted k-nearest neighbor score according to claim 6, wherein the calculation formula of the similarity R in the similarity calculation module is:
    Figure PCTCN2017080786-appb-100006
    Figure PCTCN2017080786-appb-100006
    上述公式中，Xu和Xv分别代表2个用户u和v的直播间观看向量，T代表矩阵转置，||Xu||和||Xv||分别代表Xu和Xv的模。In the above formula, X u and X v represent the live broadcast viewing vectors of two users u and v, respectively, T represents matrix transposition, and ||X u || and ||X v || represent X u and X v respectively . mold.
  10. 如权利要求6至9任一项所述的基于加权k近邻评分的直播间推荐系统，其特征在于：该系统还包括直播间推荐列表获取模块，其用于：将所述每个用户的推荐直播间形成直播间推荐列表。 The live broadcast recommendation system based on the weighted k-nearest neighbor score according to any one of claims 6 to 9, wherein the system further comprises a live broadcast recommendation list obtaining module, configured to: recommend each user The live room forms a recommended list of live rooms.
PCT/CN2017/080786 2016-08-16 2017-04-17 Weighted k-nearest-neighbor scoring-based live broadcast room recommendation method and system WO2018032790A1 (en)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201610671104.4 2016-08-16
CN201610671104.4A CN106294800A (en) 2016-08-16 2016-08-16 Method and system recommended by direct broadcasting room based on weighting k neighbour scoring

Publications (1)

Publication Number Publication Date
WO2018032790A1 true WO2018032790A1 (en) 2018-02-22

Family

ID=57671279

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2017/080786 WO2018032790A1 (en) 2016-08-16 2017-04-17 Weighted k-nearest-neighbor scoring-based live broadcast room recommendation method and system

Country Status (2)

Country Link
CN (1) CN106294800A (en)
WO (1) WO2018032790A1 (en)

Cited By (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108833935A (en) * 2018-05-25 2018-11-16 广州虎牙信息科技有限公司 A kind of direct broadcasting room recommended method, device, equipment and storage medium
CN111222055A (en) * 2020-01-13 2020-06-02 广州荔支网络技术有限公司 Audio anchor recommendation method
CN111580670A (en) * 2020-05-12 2020-08-25 黑龙江工程学院 Landscape implementing method based on virtual reality
CN112052388A (en) * 2020-08-20 2020-12-08 深思考人工智能科技(上海)有限公司 Method and system for recommending gourmet stores
CN112702618A (en) * 2020-12-16 2021-04-23 广州市千钧网络科技有限公司 Attention degree processing method, attention degree processing device, attention degree processing equipment and readable storage medium
CN112770124A (en) * 2020-12-22 2021-05-07 Oppo广东移动通信有限公司 Method and device for entering live broadcast room, storage medium and electronic equipment
CN114302152A (en) * 2021-11-17 2022-04-08 北京乐我无限科技有限责任公司 Live broadcast room recommendation method, device, equipment and storage medium
CN114697711A (en) * 2020-12-30 2022-07-01 武汉斗鱼网络科技有限公司 Anchor recommendation method and device, electronic equipment and storage medium
CN118012920A (en) * 2024-04-09 2024-05-10 青岛益生康健科技股份有限公司 User portrait label quick matching method based on big data

Families Citing this family (21)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106294800A (en) * 2016-08-16 2017-01-04 武汉斗鱼网络科技有限公司 Method and system recommended by direct broadcasting room based on weighting k neighbour scoring
CN106954086B (en) * 2017-02-28 2020-08-04 北京潘达互娱科技有限公司 Information recommendation method and device
CN107172501B (en) * 2017-03-30 2020-05-12 武汉斗鱼网络科技有限公司 Live broadcast room recommendation display method and system
CN106982381B (en) * 2017-03-31 2021-02-02 武汉斗鱼网络科技有限公司 Home page recommendation processing method and device
CN107181967A (en) * 2017-04-01 2017-09-19 北京潘达互娱科技有限公司 A kind of image display method and device
CN107172452B (en) * 2017-04-25 2020-02-18 北京潘达互娱科技有限公司 Live broadcast room recommendation method and device
CN107205178B (en) * 2017-04-25 2019-12-10 北京潘达互娱科技有限公司 Live broadcast room recommendation method and device
CN109151542B (en) * 2017-06-28 2021-07-23 武汉斗鱼网络科技有限公司 Method, device and equipment for processing illegal live broadcast room and computer readable storage medium
CN107613395B (en) * 2017-08-28 2019-11-15 武汉斗鱼网络科技有限公司 Room recommended method, system, equipment and storage medium is broadcast live
CN108156468A (en) * 2017-09-30 2018-06-12 上海掌门科技有限公司 A kind of method and apparatus for watching main broadcaster's live streaming
CN107835441B (en) * 2017-10-10 2020-01-03 武汉斗鱼网络科技有限公司 Live broadcast recommendation method, storage medium, device and system based on path prediction
CN108093274A (en) * 2018-01-05 2018-05-29 武汉斗鱼网络科技有限公司 A kind of direct broadcasting room sort method, electronic equipment and readable storage medium storing program for executing
CN110012318B (en) * 2018-01-05 2021-05-28 武汉斗鱼网络科技有限公司 Method, storage medium, device and system for determining user interest
CN108184148B (en) * 2018-01-08 2019-10-22 武汉斗鱼网络科技有限公司 A kind of method, apparatus and computer equipment of user for identification
CN108322829B (en) * 2018-03-02 2020-11-27 北京奇艺世纪科技有限公司 Personalized anchor recommendation method and device and electronic equipment
CN109218769B (en) * 2018-09-30 2021-01-01 武汉斗鱼网络科技有限公司 Recommendation method for live broadcast room and related equipment
CN109495770B (en) * 2018-11-23 2021-03-16 武汉斗鱼网络科技有限公司 Live broadcast room recommendation method, device, equipment and medium
CN109348260B (en) * 2018-12-06 2021-09-07 武汉瓯越网视有限公司 Live broadcast room recommendation method, device, equipment and medium
CN109951725B (en) * 2019-03-07 2021-06-15 武汉斗鱼鱼乐网络科技有限公司 Recommendation method for live broadcast room and related equipment
CN113159855B (en) * 2021-04-30 2023-01-13 青岛檬豆网络科技有限公司 Live broadcast recommendation method
CN114374854B (en) * 2021-12-20 2024-05-31 广西壮族自治区公众信息产业有限公司 Cloud tourism live broadcast method and system

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7454775B1 (en) * 2000-07-27 2008-11-18 Koninklijke Philips Electronics N.V. Method and apparatus for generating television program recommendations based on similarity metric
CN104504149A (en) * 2015-01-08 2015-04-08 中国联合网络通信集团有限公司 Application recommendation method and device
CN105095442A (en) * 2015-07-23 2015-11-25 海信集团有限公司 Multimedia data recommendation method and device
CN105404700A (en) * 2015-12-30 2016-03-16 山东大学 Collaborative filtering-based video program recommendation system and recommendation method
CN105574198A (en) * 2015-12-28 2016-05-11 海信集团有限公司 Column recommendation method and device
CN105808537A (en) * 2014-12-29 2016-07-27 Tcl集团股份有限公司 A Storm-based real-time recommendation method and a system therefor
CN106294800A (en) * 2016-08-16 2017-01-04 武汉斗鱼网络科技有限公司 Method and system recommended by direct broadcasting room based on weighting k neighbour scoring

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7454775B1 (en) * 2000-07-27 2008-11-18 Koninklijke Philips Electronics N.V. Method and apparatus for generating television program recommendations based on similarity metric
CN105808537A (en) * 2014-12-29 2016-07-27 Tcl集团股份有限公司 A Storm-based real-time recommendation method and a system therefor
CN104504149A (en) * 2015-01-08 2015-04-08 中国联合网络通信集团有限公司 Application recommendation method and device
CN105095442A (en) * 2015-07-23 2015-11-25 海信集团有限公司 Multimedia data recommendation method and device
CN105574198A (en) * 2015-12-28 2016-05-11 海信集团有限公司 Column recommendation method and device
CN105404700A (en) * 2015-12-30 2016-03-16 山东大学 Collaborative filtering-based video program recommendation system and recommendation method
CN106294800A (en) * 2016-08-16 2017-01-04 武汉斗鱼网络科技有限公司 Method and system recommended by direct broadcasting room based on weighting k neighbour scoring

Cited By (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108833935A (en) * 2018-05-25 2018-11-16 广州虎牙信息科技有限公司 A kind of direct broadcasting room recommended method, device, equipment and storage medium
CN111222055A (en) * 2020-01-13 2020-06-02 广州荔支网络技术有限公司 Audio anchor recommendation method
CN111580670B (en) * 2020-05-12 2023-06-30 黑龙江工程学院 Garden landscape implementation method based on virtual reality
CN111580670A (en) * 2020-05-12 2020-08-25 黑龙江工程学院 Landscape implementing method based on virtual reality
CN112052388A (en) * 2020-08-20 2020-12-08 深思考人工智能科技(上海)有限公司 Method and system for recommending gourmet stores
CN112702618B (en) * 2020-12-16 2022-12-09 广州市千钧网络科技有限公司 Attention degree processing method, attention degree processing device, attention degree processing equipment and readable storage medium
CN112702618A (en) * 2020-12-16 2021-04-23 广州市千钧网络科技有限公司 Attention degree processing method, attention degree processing device, attention degree processing equipment and readable storage medium
CN112770124A (en) * 2020-12-22 2021-05-07 Oppo广东移动通信有限公司 Method and device for entering live broadcast room, storage medium and electronic equipment
CN112770124B (en) * 2020-12-22 2023-10-31 Oppo广东移动通信有限公司 Method and device for entering live broadcast room, storage medium and electronic equipment
CN114697711A (en) * 2020-12-30 2022-07-01 武汉斗鱼网络科技有限公司 Anchor recommendation method and device, electronic equipment and storage medium
CN114697711B (en) * 2020-12-30 2024-02-20 广州财盟科技有限公司 Method and device for recommending anchor, electronic equipment and storage medium
CN114302152A (en) * 2021-11-17 2022-04-08 北京乐我无限科技有限责任公司 Live broadcast room recommendation method, device, equipment and storage medium
CN118012920A (en) * 2024-04-09 2024-05-10 青岛益生康健科技股份有限公司 User portrait label quick matching method based on big data

Also Published As

Publication number Publication date
CN106294800A (en) 2017-01-04

Similar Documents

Publication Publication Date Title
WO2018032790A1 (en) Weighted k-nearest-neighbor scoring-based live broadcast room recommendation method and system
CN104199896B (en) The video similarity of feature based classification is determined and video recommendation method
CN108322829B (en) Personalized anchor recommendation method and device and electronic equipment
JP5432243B2 (en) Send and react to media object queries
CN104935963B (en) A kind of video recommendation method based on timing driving
KR101620748B1 (en) Item recommendation method and apparatus
CN107454474B (en) A kind of television terminal program personalized recommendation method based on collaborative filtering
US11663661B2 (en) Apparatus and method for training a similarity model used to predict similarity between items
US20140344275A1 (en) Information processing apparatus, information processing method, information processing program, and recording medium
JP2013218638A (en) Content distribution system and recommendation method
CN109862431B (en) MCL-HCF algorithm-based television program mixed recommendation method
CN105208411B (en) A kind of method and device for realizing DTV target audience statistics
Lin et al. Personalized channel recommendation on live streaming platforms
US20130031093A1 (en) Information processing system, information processing method, program, and non-transitory information storage medium
CN106095974A (en) Commending system score in predicting based on network structure similarity and proposed algorithm
KR102057455B1 (en) Method for providing recommended broadcast and apparatus using the same
CN110113673A (en) A kind of barrage display methods, device and electronic equipment
CN112506977A (en) Interval intuitive fuzzy multi-attribute group decision provider selection method
US20180018395A1 (en) Landmark recommendation method and non-transitory computer-readable storage medium integrated with life behavior analysis and social network
CN105354720B (en) A method of mixed recommendation is carried out to consumption place based on visual cluster
Yi et al. A movie cold-start recommendation method optimized similarity measure
Shah et al. A hybrid based recommendation system based on clustering and association
JP5370351B2 (en) Information processing method, information processing apparatus, and information processing program
CN112948701B (en) Information recommendation device, method, equipment and storage medium
JP6856084B2 (en) Information processing device, content control device, information processing method, and program

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 17840766

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 17840766

Country of ref document: EP

Kind code of ref document: A1