WO2019041706A1 - 直播房间推荐方法及系统 - Google Patents

直播房间推荐方法及系统 Download PDF

Info

Publication number
WO2019041706A1
WO2019041706A1 PCT/CN2017/120137 CN2017120137W WO2019041706A1 WO 2019041706 A1 WO2019041706 A1 WO 2019041706A1 CN 2017120137 W CN2017120137 W CN 2017120137W WO 2019041706 A1 WO2019041706 A1 WO 2019041706A1
Authority
WO
WIPO (PCT)
Prior art keywords
live
valid
rooms
room
user
Prior art date
Application number
PCT/CN2017/120137
Other languages
English (en)
French (fr)
Inventor
王璐
张文明
陈少杰
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 WO2019041706A1 publication Critical patent/WO2019041706A1/zh

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
    • H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
    • H04N21/43—Processing of content or additional data, e.g. demultiplexing additional data from a digital video stream; Elementary client operations, e.g. monitoring of home network or synchronising decoder's clock; Client middleware
    • H04N21/442—Monitoring of processes or resources, e.g. detecting the failure of a recording device, monitoring the downstream bandwidth, the number of times a movie has been viewed, the storage space available from the internal hard disk
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
    • H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
    • H04N21/45—Management operations performed by the client for facilitating the reception of or the interaction with the content or administrating data related to the end-user or to the client device itself, e.g. learning user preferences for recommending movies, resolving scheduling conflicts
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
    • H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
    • H04N21/45—Management operations performed by the client for facilitating the reception of or the interaction with the content or administrating data related to the end-user or to the client device itself, e.g. learning user preferences for recommending movies, resolving scheduling conflicts
    • H04N21/466—Learning process for intelligent management, e.g. learning user preferences for recommending movies

Definitions

  • the present invention relates to the field of Internet live broadcast technology, and more particularly to a live room recommendation method and system.
  • the big data algorithm can be used to solve the technical problem of the above-mentioned live room recommendation.
  • a simple and feasible solution is to recommend the live broadcast room similar to the live broadcast room recently.
  • the difficulty of this solution is how to accurately calculate the similarity of the live content contained in each live broadcast room. Sex.
  • the algorithm calculates each live room based on the set.
  • the degree of similarity between the rooms is equal to the number of users viewing the two rooms at the same time divided by the number of users who have viewed at least one of the rooms.
  • the disadvantage of this method is that only the viewing users of two rooms are considered, and the viewing of other rooms by these users is not considered, only part of the information is utilized, and the similarity measure is relatively one-sided.
  • the present invention provides a live room recommendation method for overcoming the above problems or at least partially solving the above problems.
  • a live room recommendation method including:
  • Step 1 selecting a valid user and a valid live room from the current target range user and the target range live room in the target Internet live broadcast platform;
  • Step 2 Calculate a similarity value between each of the valid live rooms according to a valid user set shared by each of the active live rooms between the two preset historical time periods;
  • Step 3 Obtain each historical live broadcast room viewed in the preset time period before the current time of any valid user, and recommend the corresponding effective live broadcast room with the highest similarity value to each of the historical effective live broadcast rooms to any valid user. .
  • step 1 further includes:
  • Selecting a live room in which the number of valid users in the second preset historical time period reaches a preset number is an effective live room.
  • step 2 further includes:
  • i and j are any two different valid live rooms;
  • U i and U j are valid user sets for viewing live live rooms i and j respectively during the first preset historical time period;
  • I u and I v are respectively a set of valid live rooms viewed by user u and user v;
  • step S22 further includes: S ⁇ i, j> is inversely proportional to the
  • the step 1 further includes:
  • step S22 further includes:
  • step S22 further includes:
  • ⁇ is a constant; a penalty factor for a valid user u; A penalty factor for an effective live broadcast, where ⁇ is a constant greater than zero.
  • a live room recommendation system including a screening module, an acquisition module, and a recommendation module:
  • the screening module is electrically connected to the obtaining module, and is configured to filter out a valid user and a valid live room from the current target range user and the target range live room in the target Internet live broadcast platform;
  • the obtaining module is electrically connected to the screening module and the recommendation module, respectively, for calculating each of the valid user sets that are jointly owned by each of the active live rooms during the first preset historical time period.
  • the recommendation module is electrically connected to the acquisition module, and is configured to obtain each historical live room viewed in the preset time period before the current time of any valid user, and recommend to the valid user to be valid with the history.
  • an apparatus for recommending a live room including:
  • At least one processor At least one processor
  • At least one memory communicatively coupled to the processor, wherein:
  • the memory stores program instructions executable by the processor, the processor invoking the program instructions to perform the method of any of the above.
  • a non-transitory computer readable storage medium stores computer instructions that cause the computer to perform any of the above Said method.
  • the live room recommendation method and system proposed by the present application can calculate the degree of association of different live rooms on the basis of considering the number of live rooms shared by different users, and can accurately measure the similarity between live rooms, thereby being accurate and effective.
  • the recommendation of the live room is performed; the present invention has the beneficial effect of solving the one-sided problem of the similarity measure considerations appearing in the related art algorithms.
  • FIG. 1 is a schematic overall flow chart of a method for recommending a live room according to an embodiment of the present invention
  • FIG. 2 is a schematic diagram of an overall framework of a live room recommendation system according to an embodiment of the present invention
  • FIG. 3 is a schematic structural diagram of an apparatus for recommending a live room according to an embodiment of the present invention.
  • the recommendation results are generated by mining the relationship between the recommended items by the algorithm, and the invention effectively implements the push between the effective live rooms based on the recommended idea.
  • Similar live room A live room calculated by an algorithm similar to the one that users frequently watch live broadcasts.
  • the following specific embodiments of the present invention propose at least one live room recommendation method and system.
  • FIG. 1 is a general flow chart showing a method for recommending a live room in a specific embodiment of the present invention. In general, including:
  • Step 1 selecting a valid user and a valid live room from the current target range user and the target range live room in the target Internet live broadcast platform;
  • Step 2 Calculate a similarity value between each of the valid live rooms according to a valid user set shared by each of the active live rooms between the two preset historical time periods;
  • Step 3 Obtain each historical live broadcast room viewed in the preset time period before the current time of any valid user, and recommend the corresponding effective live broadcast room with the highest similarity value to each of the historical effective live broadcast rooms to any valid user. .
  • a method for recommending a live room the step 1 further includes:
  • Selecting a live room in which the number of valid users in the second preset historical time period reaches a preset number is an effective live room.
  • a live room recommendation method the step 2 further includes:
  • i and j are any two different valid live rooms;
  • U i and U j are valid user sets for viewing live live rooms i and j respectively during the first preset historical time period;
  • I u and I v are respectively a set of valid live rooms viewed by user u and user v;
  • a live room recommendation method further includes: S ⁇ i,j> and the
  • a method for recommending a live room the step 1 further includes:
  • a live room recommendation method further includes:
  • a live room recommendation method further includes:
  • ⁇ is a constant; a penalty factor for a valid user u; A penalty factor for an effective live broadcast, where ⁇ is a constant greater than zero.
  • the recommended method for live broadcast based on the triangular graph structure proposed by the present invention is calculated based on the "triangle map structure recommendation" algorithm.
  • the "Triangle Structure Recommendation” algorithm is intended to calculate the similarity between the live rooms.
  • the design idea of the triangle structure recommendation algorithm is as follows.
  • the scheme defines a "triangular structure" in the algorithm. If both users a and b have viewed the same live room r at the same time, then (a, b, r) can form a stable triangular structure. If the number of triangle structures formed between two users a and b is more, the weaker the relationship between a and b is; in other words, if the triangle structure formed between two users a and b The less, the stronger the association between a and b.
  • S ⁇ i,j> represents the similarity value between the live room i and the live room j.
  • U i represents a set of users who have watched the live room i
  • U j represents a set of users who have watched the live room j.
  • I u represents the live broadcast collection that the user u has seen
  • I v represents the live broadcast collection that the user v has seen.
  • ⁇ is a constant and is generally set to 1 in the algorithm.
  • the user and the room are cleaned, and the blacklisted users with the brushing behavior and the rooms with a small number of monthly views are removed; the viewing behavior of the user watching for more than one minute is included in the statistical range as an effective viewing behavior.
  • the number of rooms co-viewed between the two users after the deletion is calculated; the similarity between the live broadcasts is calculated by the above-mentioned "triangular structure recommendation” algorithm.
  • the "triangular structure recommendation” algorithm uses the "collective user history viewing data" of the entire network user, so the individual data noise is not Will affect the measurement results of the algorithm, anti-noise ability is very strong.
  • cosine Distance algorithm the algorithm takes each user's viewing behavior on the room as a dimension of the vector The similarity between rooms is calculated by the cosine formula.
  • the disadvantage of this method is that it is easy to produce "Harry Potter” phenomenon.
  • the head popular anchor will appear in the user's recommendation set; the anti-noise ability is not strong, and often there will be some
  • the wrong metric introduces a penalty factor in the algorithm formula, punishing active users with too many behaviors and popular live broadcast rooms to improve accuracy.
  • This embodiment proposes an improved "triangle map structure recommendation” algorithm formula as follows:
  • the similarity between two pairs in the live broadcast can be calculated:
  • the similarity of the rooms I1 and I2 is the highest, so the room I2 can be recommended to the user U1 who has seen I1, and the I1 is recommended to the user U3 who has seen I2.
  • the method of the present invention can accurately measure the similarity between the live broadcasts by utilizing the triangular structure relationship formed by the user and the commodity, and solves the problem that the anti-noise ability is similar in the similar algorithm.
  • FIG. 2 an overall frame diagram of a live room recommendation system in accordance with one embodiment of the present invention.
  • the screening module A1, the acquisition module A2, and the recommendation module A3 are included:
  • the screening module A1 is electrically connected to the obtaining module A2, and is configured to filter out a valid user and a valid live room from the current target range user and the target range live room in the target Internet live broadcast platform;
  • the obtaining module A2 is electrically connected to the screening module A1 and the recommendation module A3, respectively, and is configured to calculate a valid user set that is jointly owned by each of the active live rooms in the first preset historical time period.
  • the recommendation module A3 is electrically connected to the acquisition module A2, and is configured to acquire each historical live broadcast room viewed in the preset time period before the current time of any valid user, and recommend and describe the valid user to each of the valid users.
  • a live room recommendation system is further configured to: select to continuously view any fixed live room, the user whose length exceeds the preset length, and/or the non-blacklist user is valid. The user selects a live room that effectively watches the number of users in the second preset historical time period to a preset number of live broadcast rooms.
  • a live room recommendation system the obtaining module is further configured to:
  • i and j are any two different valid live rooms;
  • U i and U j are valid user sets for viewing live live rooms i and j respectively during the first preset historical time period;
  • I u and I v are respectively a set of valid live rooms viewed by user u and user v;
  • a live room recommendation system the obtaining module is further configured to: S ⁇ i,j> and the
  • a live room recommendation system is further configured to: screen out valid users and effective live rooms from all current online users and all live rooms in the target Internet live broadcast platform. .
  • a live room recommendation system the obtaining module is further configured to:
  • a live room recommendation system the obtaining module is further configured to:
  • ⁇ is a constant; a penalty factor for a valid user u; A penalty factor for an effective live broadcast, where ⁇ is a constant greater than zero.
  • FIG. 3 is a structural block diagram of an apparatus for a live room recommendation method according to an embodiment of the present application.
  • the device of the live room recommendation method includes: a processor 301, a memory 302, and a bus 303;
  • the processor 301 and the memory 302 complete communication with each other through the bus 303;
  • the processor 301 is configured to invoke the program instructions in the memory 302 to perform the methods provided by the foregoing method embodiments, for example, including: filtering out the current target range user and the target range live room in the target Internet live broadcast platform. a valid user and a valid live room; calculating a similarity value between each of the valid live rooms based on a set of valid users shared by the active live rooms between the two preset historical time periods; Each historical live live room viewed within a preset time period before the current time of any valid user, and the corresponding valid live broadcast room having the highest similarity value to each of the historical effective live broadcast rooms is recommended to the valid user.
  • the embodiment discloses a computer program product comprising a computer program stored on a non-transitory computer readable storage medium, the computer program comprising program instructions, when the program instructions are executed by a computer, the computer
  • the method provided by the foregoing method embodiments can be performed, for example,: selecting a valid user and a valid live room from the current target range user and the target range live room in the target Internet live broadcast platform; and based on each of the effective live rooms Calculating the similarity value between each of the valid live rooms in a set of valid users shared by the first preset historical time period; obtaining the history viewed within the preset time period before the current time of any valid user
  • the live room is effectively live, and the corresponding valid live room with the highest similarity value to each of the historical effective live rooms is recommended to the any valid user.
  • the embodiment provides a non-transitory computer readable storage medium, the non-transitory computer readable storage medium storing computer instructions, the computer instructions causing the computer to perform the methods provided by the foregoing method embodiments, including, for example, : filtering valid users and valid live rooms from the current target range user and the target range live room in the target Internet live broadcast platform; and validly sharing between the two active live rooms in the first preset historical time period
  • the user set calculates a similarity value between each of the valid live rooms; obtains each historical live room viewed in the preset time period before the current time of any valid user, and recommends to the valid user
  • the corresponding effective live room with the highest similarity value in each historical live broadcast room is described.
  • the foregoing program may be stored in a computer readable storage medium, and the program is executed when executed.
  • the foregoing steps include the steps of the foregoing method embodiments; and the foregoing storage medium includes: a medium that can store program codes, such as a ROM, a RAM, a magnetic disk, or an optical disk.
  • the embodiments of the device and the like for the live room recommendation method described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as the unit may or may not be Physical units can be located in one place or distributed to multiple network elements. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the embodiment. Those of ordinary skill in the art can understand and implement without deliberate labor.

Landscapes

  • Engineering & Computer Science (AREA)
  • Databases & Information Systems (AREA)
  • Multimedia (AREA)
  • Signal Processing (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
  • Two-Way Televisions, Distribution Of Moving Picture Or The Like (AREA)
  • Testing, Inspecting, Measuring Of Stereoscopic Televisions And Televisions (AREA)

Abstract

本发明提供直播房间推荐方法及系统，所述方法在考虑不同用户间所共同观看直播房间数量的基础上，计算不同直播房间的关联程度，根据各直播房间质检的关联程度大小进行直播房间的推荐。本发明能够准确地度量各直播房间之间相似度，从而精准有效的进行直播房间的推荐；具有解决现有直播房间推荐技术方案由于现有相关算法引起的相似性度量考虑因素片面问题的有益效果。

Description

直播房间推荐方法及系统
交叉引用
本申请引用于2017年08月28日提交的专利名称为“直播房间推荐方法及系统”的第2017107496872号中国专利申请，其通过引用被全部并入本申请。
技术领域
本发明涉及互联网直播技术领域，更具体地，涉及直播房间推荐方法及系统。
背景技术
随着互联网直播平台上内容地不断丰富与繁杂的今天，如何给当前用户准确的推荐其感兴趣的直播间是一个众多互联网直播平台所关注的问题。
现有技术中，可以利用大数据算法解决上述直播房间推荐的技术问题。在直播间推荐的众多大数据算法解决方案中，一个简单可行的方案是给用户推荐最近观看过直播间类似的直播间，该方案的难点在于如何准确地计算各直播间所包含直播内容的相似性。
现有技术中的直播房间推荐方案中，有以下一种度量直播房间相似度的杰卡德系数算法(Jaccard’s Coefficient)，该算法在直播房间推荐中存在如下问题：该算法基于集合计算各直播房间之间的相似度，房间之间的相似度等于同时观看两个房间的用户数除以至少观看了其中一个房间的用户数。该方法的缺点在于仅仅考虑了两个房间的观看用户，未考虑这些用户对其他房间的观看情况，利用的只是部分信息，相似性度量相对而言比较片面。
现有技术亟需一种可以解决现有技术中所存在相似性度量考虑因素片面问题的直播房间推荐方法。
发明内容
本发明为克服上述问题或者至少部分地解决上述问题，提供直播房间推荐方法。
根据本发明的一个方面，提供一种直播房间推荐方法，包括：
步骤1，从目标互联网直播平台中当前目标范围用户和目标范围直播房间中筛选出有效用户和有效直播房间；
步骤2，基于各所述有效直播房间两两之间在第一预设历史时间段内共同拥有的有效用户集合计算各所述有效直播房间两两之间的相似度值；
步骤3，获取任一有效用户当前时刻之前预设时长内所观看过的各历史有效直播房间，向所述任一有效用户推荐与所述各历史有效直播房间相似度值最高的相应有效直播房间。
进一步，所述步骤1进一步包括：
选取连续观看任一固定直播房间时长超过预设长的用户和/或非黑名单用户为有效用户；
选取第二预设历史时间段内有效观看用户人数达到预设数量的直播房间为有效直播房间。
进一步，所述步骤2进一步包括：
S21，获取各所述有效直播房间两两之间在第一预设历史时间段内共同拥有的有效用户集合U i∩U j；获取所述U i∩U j中各有效用户两两之间在第一预设历史时间段内共同观看过的有效直播房间数量|I u∩I v|；
其中，i和j为任意两个不同的有效直播房间；U i和U j分别为在第一预设历史时间段内观看过有效直播房间i和j的有效用户集合；u∈U i∩U j,v∈U i∩U j；I u和I v分别为用户u和用户v观看过的有效直播房间集合；
S22，基于所述|I u∩I v|计算各所述有效直播房间两两之间的相似度值S<i,j>。
进一步，所述步骤S22进一步包括：S<i,j>与所述有效直播房间i和j之间的所述|I u∩I v|成反比关系。
所述步骤1进一步包括：
从目标互联网直播平台中当前所有在线用户和所有正在直播房间中筛选出有效用户和有效直播房间。
进一步，所述步骤S22进一步包括：
Figure PCTCN2017120137-appb-000001
其中，α为常数。
进一步，所述步骤S22进一步包括：
Figure PCTCN2017120137-appb-000002
其中，α为常数；
Figure PCTCN2017120137-appb-000003
为对有效用户u的惩罚因子；
Figure PCTCN2017120137-appb-000004
为对有效直播间j的惩罚因子，其中β是一个大于零的常数。
根据本发明另一方面，提供一种直播房间推荐系统，包括筛选模块、获取模块和推荐模块：
所述筛选模块与所述获取模块电连接，用于从目标互联网直播平台中当前目标范围用户和目标范围直播房间中筛选出有效用户和有效直播房间；
所述获取模块分别与所述筛选模块和所述推荐模块电连接，用于基于各所述有效直播房间两两之间在第一预设历史时间段内共同拥有的有效用户集合计算各所述有效直播房间两两之间的相似度值S<i,j>；
所述推荐模块与所述获取模块电连接，用于获取任一有效用户当前时刻之前预设时长内所观看过的各历史有效直播房间，向所述任一有效用户推荐与所述各历史有效直播房间相似度值最高的相应有效直播房间。
根据本发明另一方面，提供一种直播房间推荐方法的设备，包括：
至少一个处理器；以及
与所述处理器通信连接的至少一个存储器，其中：
所述存储器存储有可被所述处理器执行的程序指令，所述处理器调用所述程序指令能够执行如上述任一所述的方法。
根据本发明另一方面，提供一种非暂态计算机可读存储介质，其特征在于，所述非暂态计算机可读存储介质存储计算机指令，所述计算机指令使所述计算机执行如上述任一所述的方法。
本申请提出的直播房间推荐方法及系统，在考虑不同用户间所共同观看直播房间数量的基础上，计算不同直播房间的关联程度，能准确地度量各直播房间之间相似度，从而精准有效的进行直播房间的推荐；本发明具有解决了现有技术相关算法中出现的相似性度量考虑因素片面问题的有益效果。
附图说明
图1为根据本发明实施例一种直播房间推荐方法的整体流程示意图；
图2为根据本发明实施例又一种直播房间推荐系统的整体框架示意图；
图3为根据本发明实施例一种直播房间推荐方法的设备的结构框架示意图。
具体实施方式
下面结合附图和实施例，对本发明的具体实施方式作进一步详细描述。以下实施例用于说明本发明，但不用来限制本发明的范围。
名词解释：
基于物品的推荐：通过算法挖掘推荐物品之间的关系生成推荐结果，本发明基于该推荐思路，有效实现了有效直播房间之间推送。
相似直播间：通过算法计算出的与用户最近经常观看直播间类似的直播间。
哈利波特现象：推荐算法中在很多用户推荐结果中都会出现热门物品的现象。
在基于物品的推荐方案中，存在背景技术中所提到的a)余弦距离算法(Cosine Distance)和b)杰卡德系数算法(Jaccard’s Coefficient)，这些算法在其他推荐场景中广泛使用，但在直播房间推荐中都相应存在“哈利波特”现象和相似性度量考虑片面等问题。
为了改善上述算法的缺点，本发明下述具体实施例提出至少一种直播房间推荐方法及系统。
如图1，示出本发明的一个具体实施例中，一种直播房间推荐方法的整体流程图。总体上，包括：
步骤1，从目标互联网直播平台中当前目标范围用户和目标范围直播房间中筛选出有效用户和有效直播房间；
步骤2，基于各所述有效直播房间两两之间在第一预设历史时间段内共同拥有的有效用户集合计算各所述有效直播房间两两之间的相似度值；
步骤3，获取任一有效用户当前时刻之前预设时长内所观看过的各历史有效直播房间，向所述任一有效用户推荐与所述各历史有效直播房间相似度值最高的相应有效直播房间。
在本发明上述具体实施例的基础上，一种直播房间推荐方法，所述步骤1进一步包括：
选取连续观看任一固定直播房间时长超过预设长的用户和/或非黑名单用户为有效用户；
选取第二预设历史时间段内有效观看用户人数达到预设数量的直播房间为有效直播房间。
在本发明任一上述具体实施例的基础上，一种直播房间推荐方法，所述步骤2进一步包括：
S21，获取各所述有效直播房间两两之间在第一预设历史时间段内共同拥有的有效用户集合U i∩U j；获取所述U i∩U j中各有效用户两两之间在第一预设历史时间段内共同观看过的有效直播房间数量|I u∩I v|；
其中，i和j为任意两个不同的有效直播房间；U i和U j分别为在第一预设历史时间段内观看过有效直播房间i和j的有效用户集合；u∈U i∩U j,v∈U i∩U j；I u和I v分别为用户u和用户v观看过的有效直播房间集合；
S22，基于所述|I u∩I v|计算各所述有效直播房间两两之间的相似 度值S<i,j>。
在本发明任一上述具体实施例的基础上，一种直播房间推荐方法，所述步骤S22进一步包括：S<i,j>与所述有效直播房间i和j之间的所述|I u∩I v|成反比关系。
在本发明任一上述具体实施例的基础上，一种直播房间推荐方法，所述步骤1进一步包括：
从目标互联网直播平台中当前所有在线用户和所有正在直播房间中筛选出有效用户和有效直播房间。
在本发明任一上述具体实施例的基础上，一种直播房间推荐方法，所述步骤S22进一步包括：
Figure PCTCN2017120137-appb-000005
其中，α为常数。
在本发明任一上述具体实施例的基础上，一种直播房间推荐方法，所述步骤S22进一步包括：
Figure PCTCN2017120137-appb-000006
其中，α为常数；
Figure PCTCN2017120137-appb-000007
为对有效用户u的惩罚因子；
Figure PCTCN2017120137-appb-000008
为对有效直播间j的惩罚因子，其中β是一个大于零的常数。
在本发明任一上述具体实施例的基础上，一种直播房间推荐方法，接下来围绕下述几个方面对本发明技术方案进行展开说明。
本发明提出的基于三角图结构的直播间推荐方法是基于“三角图结构推荐”算法所进行计算的。
所述“三角图结构推荐”算法意在计算所述各直播间之间的相似度。所述三角图结构推荐”算法的设计思想如下。
如果多个用户在观看了同一个直播间s的同时，都只另外共同观看了同一个直播间r，那么可以得到r和s为强关联度的结论。根据上述思想，本方案在算法中定义了一种“三角图结构”，如果用户a和b都同时观看过同一直播房间r，那么(a,b,r)就可以构成一个稳定的三角图结构；如果两个用户a和b之间构成的三角图结构越多，那么所述a和b之间的关联关系越弱；反而言之，如果两个用户a和b之间构成的三角图结构越少，那么所述a和b之间的关联关系越强。
基于上述推理，可以设计了如下“三角图结构推荐”算法的核心公式：
Figure PCTCN2017120137-appb-000009
其中：
S<i,j>表示直播间i和直播间j的相似度值。
U i表示观看过直播间i的用户集合，U j表示观看过直播间j的用户集合。
I u表示用户u看过的直播间集合，I v表示用户v看过的直播间集合。
α是一个常数，在算法中一般设置为1。
在方案具体实施例中，本推荐方案的实施步骤如下：
首先，对用户和房间进行清洗，除去有刷量行为的黑名单用户和月观看人数极少的房间；将用户观看时长超过1分钟的观看行为作为有效的观看行为纳入统计范围中。
其次，计算删选过后的用户两两之间共同观看的房间数；利用前 述“三角图结构推荐”算法计算直播间之间的相似度。
最后，根据用户近期观看的房间推荐与其相似度最大的若干个房间作为推荐集。
在本发明上述任一实施例的基础上，一种直播房间推荐方法，上述实施例中“三角图结构推荐”算法运用了全网用户的“集体用户历史观看数据”，因此个别的数据噪声不会影响到算法的度量结果，抗噪音能力非常强。
在本发明具体实施例中，同时为了解决现有技术余弦距离算法中“哈利波特”现象(余弦距离算法(Cosine Distance)：该算法将每个用户对房间的观看行为作为向量的一个维度，房间之间的相似度采用余弦公式计算，该方法的缺点在于易产生“哈利波特”现象，在用户的推荐集中都会出现头部热门主播；抗噪声的能力不强，经常会出现一些错误的度量的)，在算法公式中引入惩罚因子，惩罚那些行为过多的活跃用户和热门的直播间，提高准确率。本具体实施例提出一种改进的“三角图结构推荐”算法公式如下：
Figure PCTCN2017120137-appb-000010
其中：
Figure PCTCN2017120137-appb-000011
为对用户u的惩罚因子，同理，
Figure PCTCN2017120137-appb-000012
为对直播间j的惩罚因子，其中β是一个大于零的常数；其中，在计算i的相似房间时，因为只需要对计算的房间j作惩罚，所以对i不作惩罚。s<i,j>与s<j,i>不对称，是因为同一个i对i惩罚的话都是一样，写不写从结果上是一样的。
在本发明上述任一实施例的基础上，一种直播房间推荐方法的具体应用举例。
假设现在有五位用户，分别记为U1、U2、U3、U4、U5；有三个直播间，分别记为I1、I2、I3。每位用户看过的直播间集合如下：
U1：{I1}
U2：{I1、I2、I3}
U3：{I2、I3}
U4：{I1、I2}
U5：{I1、I2}
根据基本的核心算法公式可以计算直播间两两的相似度：
S<I1,I2>＝1/(1+2)+1/(1+2)+1/(1+2)＝1
S<I1,I3>＝0
S<I2,I3>＝1/(1+2)＝1/3
根据相似度计算的结果，房间I1和I2的相似度最高，因此可以给看过I1的用户U1推荐房间I2，给看过I2的用户U3推荐I1。
综上述所有实施例所述，本发明所述方法通过利用用户和商品形成的三角结构图关系，能准确地度量直播间之间相似度，解决了类似算法中出现的抗噪能力不强和哈利波特现象的问题。
如图2，示出本发明的一个具体实施例中，一种直播房间推荐系统的整体框架图。总体上，包括筛选模块A1、获取模块A2和推荐模块A3：
所述筛选模块A1与所述获取模块A2电连接，用于从目标互联网直播平台中当前目标范围用户和目标范围直播房间中筛选出有效用户和有效直播房间；
所述获取模块A2分别与所述筛选模块A1和所述推荐模块A3电 连接，用于基于各所述有效直播房间两两之间在第一预设历史时间段内共同拥有的有效用户集合计算各所述有效直播房间两两之间的相似度值S<i,j>；
所述推荐模块A3与所述获取模块A2电连接，用于获取任一有效用户当前时刻之前预设时长内所观看过的各历史有效直播房间，向所述任一有效用户推荐与所述各历史有效直播房间相似度值最高的相应有效直播房间。
在上述任一具体实施例的基础上，一种直播房间推荐系统，所述筛选模块进一步用于：选取连续观看任一固定直播房间时长超过预设长的用户和/或非黑名单用户为有效用户；选取第二预设历史时间段内有效观看用户人数达到预设数量的直播房间为有效直播房间。
在上述任一具体实施例的基础上，一种直播房间推荐系统，所述获取模块进一步用于：
获取各所述有效直播房间两两之间在第一预设历史时间段内共同拥有的有效用户集合U i∩U j；获取所述U i∩U j中各有效用户两两之间在第一预设历史时间段内共同观看过的有效直播房间数量|I u∩I v|；
其中，i和j为任意两个不同的有效直播房间；U i和U j分别为在第一预设历史时间段内观看过有效直播房间i和j的有效用户集合；u∈U i∩U j,v∈U i∩U j；I u和I v分别为用户u和用户v观看过的有效直播房间集合；
基于所述|I u∩I v|计算各所述有效直播房间两两之间的相似度值S<i,j>。
在上述任一具体实施例的基础上，一种直播房间推荐系统，所述获取模块进一步用于：S<i,j>与所述有效直播房间i和j之间的所述|I u∩I v|成反比关系。
在上述任一具体实施例的基础上，一种直播房间推荐系统，所述筛选模块进一步用于：从目标互联网直播平台中当前所有在线用户和所有正在直播房间中筛选出有效用户和有效直播房间。
在上述任一具体实施例的基础上，一种直播房间推荐系统，所述获取模块进一步用于：
Figure PCTCN2017120137-appb-000013
其中，α为常数
在上述任一具体实施例的基础上，一种直播房间推荐系统，所述获取模块进一步用于：
Figure PCTCN2017120137-appb-000014
其中，α为常数；
Figure PCTCN2017120137-appb-000015
为对有效用户u的惩罚因子；
Figure PCTCN2017120137-appb-000016
为对有效直播间j的惩罚因子，其中β是一个大于零的常数。
图3示出本申请实施例的直播房间推荐方法的设备的结构框图。
参照图3，所述直播房间推荐方法的设备，包括：处理器(processor)301、存储器(memory)302和总线303；
其中，
所述处理器301和存储器302通过所述总线303完成相互间的通信；
所述处理器301用于调用所述存储器302中的程序指令，以执行上述各方法实施例所提供的方法，例如包括：从目标互联网直播平台中当前目标范围用户和目标范围直播房间中筛选出有效用户和有效直 播房间；基于各所述有效直播房间两两之间在第一预设历史时间段内共同拥有的有效用户集合计算各所述有效直播房间两两之间的相似度值；获取任一有效用户当前时刻之前预设时长内所观看过的各历史有效直播房间，向所述任一有效用户推荐与所述各历史有效直播房间相似度值最高的相应有效直播房间。
本实施例公开一种计算机程序产品，所述计算机程序产品包括存储在非暂态计算机可读存储介质上的计算机程序，所述计算机程序包括程序指令，当所述程序指令被计算机执行时，计算机能够执行上述各方法实施例所提供的方法，例如包括：从目标互联网直播平台中当前目标范围用户和目标范围直播房间中筛选出有效用户和有效直播房间；基于各所述有效直播房间两两之间在第一预设历史时间段内共同拥有的有效用户集合计算各所述有效直播房间两两之间的相似度值；获取任一有效用户当前时刻之前预设时长内所观看过的各历史有效直播房间，向所述任一有效用户推荐与所述各历史有效直播房间相似度值最高的相应有效直播房间。
本实施例提供一种非暂态计算机可读存储介质，所述非暂态计算机可读存储介质存储计算机指令，所述计算机指令使所述计算机执行上述各方法实施例所提供的方法，例如包括：从目标互联网直播平台中当前目标范围用户和目标范围直播房间中筛选出有效用户和有效直播房间；基于各所述有效直播房间两两之间在第一预设历史时间段内共同拥有的有效用户集合计算各所述有效直播房间两两之间的相似度值；获取任一有效用户当前时刻之前预设时长内所观看过的各历史有效直播房间，向所述任一有效用户推荐与所述各历史有效直播房间相似度值最高的相应有效直播房间。
本领域普通技术人员可以理解：实现上述方法实施例的全部或部分步骤可以通过程序指令相关的硬件来完成，前述的程序可以存储于一计算机可读取存储介质中，该程序在执行时，执行包括上述方法实 施例的步骤；而前述的存储介质包括：ROM、RAM、磁碟或者光盘等各种可以存储程序代码的介质。
以上所描述的直播房间推荐方法的设备等实施例仅仅是示意性的，其中所述作为分离部件说明的单元可以是或者也可以不是物理上分开的，作为单元显示的部件可以是或者也可以不是物理单元，即可以位于一个地方，或者也可以分布到多个网络单元上。可以根据实际的需要选择其中的部分或者全部模块来实现本实施例方案的目的。本领域普通技术人员在不付出创造性的劳动的情况下，即可以理解并实施。
通过以上的实施方式的描述，本领域的技术人员可以清楚地了解到各实施方式可借助软件加必需的通用硬件平台的方式来实现，当然也可以通过硬件。基于这样的理解，上述技术方案本质上或者说对现有技术做出贡献的部分可以以软件产品的形式体现出来，该计算机软件产品可以存储在计算机可读存储介质中，如ROM/RAM、磁碟、光盘等，包括若干指令用以使得一台计算机设备(可以是个人计算机，服务器，或者网络设备等)执行各个实施例或者实施例的某些部分所述的方法。
最后，本申请的方法仅为较佳的实施方案，并非用于限定本发明的保护范围。凡在本发明的精神和原则之内，所作的任何修改、等同替换、改进等，均应包含在本发明的保护范围之内。

Claims (11)

  1. 一种直播房间推荐方法，其特征在于，包括：
    步骤1，从目标互联网直播平台中当前目标范围用户和目标范围直播房间中筛选出有效用户和有效直播房间；
    步骤2，基于各所述有效直播房间两两之间在第一预设历史时间段内共同拥有的有效用户集合计算各所述有效直播房间两两之间的相似度值；
    步骤3，获取任一有效用户当前时刻之前预设时长内所观看过的各历史有效直播房间，向所述任一有效用户推荐与所述各历史有效直播房间相似度值最高的相应有效直播房间。
  2. 如权利要求1所述的方法，其特征在于，所述步骤1进一步包括：
    选取连续观看任一固定直播房间时长超过预设长的用户和/或非黑名单用户为有效用户；
    选取第二预设历史时间段内有效观看用户人数达到预设数量的直播房间为有效直播房间。
  3. 如权利要求1所述的方法，其特征在于，所述步骤2进一步包括：
    S21，获取各所述有效直播房间两两之间在第一预设历史时间段内共同拥有的有效用户集合U i∩U j；获取所述U i∩U j中各有效用户两两之间在第一预设历史时间段内共同观看过的有效直播房间数量|I u∩I v|；
    其中，i和j为任意两个不同的有效直播房间；U i和U j分别为在第一预设历史时间段内观看过有效直播房间i和j的有效用户集合；u∈U i∩U j,v∈U i∩U j；I u和I v分别为用户u和用户v观看过的有效直播房间集合；
    S22，基于所述|I u∩I v|计算各所述有效直播房间两两之间的相似度值S<i,j>。
  4. 如权利要求3所述的方法，其特征在于，所述步骤S22进一步包括：S<i,j>与所述有效直播房间i和j之间的所述|I u∩I v|成反比关系。
  5. 如权利要求1所述的方法，其特征在于，所述步骤1进一步包括：
    从目标互联网直播平台中当前所有在线用户和所有正在直播房间中筛选出有效用户和有效直播房间。
  6. 如权利要求4所述的方法，其特征在于，所述步骤S22进一步包括：
    Figure PCTCN2017120137-appb-100001
    其中，α为常数
  7. 如权利要求4所述的方法，其特征在于，所述步骤S22进一步包括：
    Figure PCTCN2017120137-appb-100002
    其中，α为常数；
    Figure PCTCN2017120137-appb-100003
    为对有效用户u的惩罚因子；
    Figure PCTCN2017120137-appb-100004
    为对有效直播间j的惩罚因子，其中β是一个大于零的常数。
  8. 一种直播房间推荐系统，其特征在于，包括筛选模块、获取模块和推荐模块：
    所述筛选模块与所述获取模块电连接，用于从目标互联网直播平台中当前目标范围用户和目标范围直播房间中筛选出有效用户和有效直播房间；
    所述获取模块分别与所述筛选模块和所述推荐模块电连接，用于基于各所述有效直播房间两两之间在第一预设历史时间段内共同拥有的有效用户集合计算各所述有效直播房间两两之间的相似度值S<i,j>；
    所述推荐模块与所述获取模块电连接，用于获取任一有效用户当前时刻之前预设时长内所观看过的各历史有效直播房间，向所述任一有效用户推荐与所述各历史有效直播房间相似度值最高的相应有效直播房间。。
  9. 一种直播房间推荐方法的设备，其特征在于，包括：
    至少一个处理器；以及
    与所述处理器通信连接的至少一个存储器，其中：
    所述存储器存储有可被所述处理器执行的程序指令，所述处理器调用所述程序指令能够执行如权利要求1至7任一所述的方法。
  10. 一种非暂态计算机可读存储介质，其特征在于，所述非暂态计算机可读存储介质存储计算机指令，所述计算机指令使所述计算机执行如权利要求1至7任一所述的方法。
  11. 一种计算机程序产品，其特征在于，所述计算机程序产品包括存储在非暂态计算机可读存储介质上的计算机程序，所述计算机程序包括程序指令，当所述程序指令被计算机执行时，使所述计算机执行如权利要求1至7任一所述的方法。
PCT/CN2017/120137 2017-08-28 2017-12-29 直播房间推荐方法及系统 WO2019041706A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201710749687.2A CN107613395B (zh) 2017-08-28 2017-08-28 直播房间推荐方法、系统、设备及存储介质
CN201710749687.2 2017-08-28

Publications (1)

Publication Number Publication Date
WO2019041706A1 true WO2019041706A1 (zh) 2019-03-07

Family

ID=61056264

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2017/120137 WO2019041706A1 (zh) 2017-08-28 2017-12-29 直播房间推荐方法及系统

Country Status (2)

Country Link
CN (1) CN107613395B (zh)
WO (1) WO2019041706A1 (zh)

Families Citing this family (15)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108600792B (zh) * 2018-04-02 2020-08-04 武汉斗鱼网络科技有限公司 一种相似度度量方法、装置、设备及存储介质
CN108777801B (zh) * 2018-05-30 2020-08-21 广州虎牙信息科技有限公司 优质主播用户挖掘方法、装置及计算机存储介质、服务器
CN109086813B (zh) * 2018-07-23 2021-05-14 广州虎牙信息科技有限公司 主播相似度的确定方法、装置、设备及存储介质
CN109086822B (zh) * 2018-08-01 2021-04-06 广州虎牙信息科技有限公司 一种主播用户分类方法、装置、设备及存储介质
CN108989893B (zh) * 2018-08-21 2020-12-15 武汉斗鱼网络科技有限公司 一种直播间推荐方法、装置、终端和存储介质
CN109286821B (zh) * 2018-09-30 2021-10-15 武汉斗鱼网络科技有限公司 一种直播间推荐方法、装置、服务器及存储介质
CN109413459B (zh) * 2018-09-30 2020-10-16 武汉斗鱼网络科技有限公司 一种直播平台中用户的推荐方法以及相关设备
CN111083503B (zh) * 2018-10-22 2022-02-22 武汉斗鱼网络科技有限公司 对直播间计算相似度的方法、装置、设备和存储介质
CN109462778B (zh) * 2018-11-15 2020-10-16 武汉斗鱼网络科技有限公司 一种直播标识的推荐方法以及相关设备
CN109587530B (zh) * 2018-11-22 2021-06-08 广州虎牙信息科技有限公司 一种数据处理方法、装置、终端设备和存储介质
CN109977264B (zh) * 2019-02-19 2022-09-09 武汉斗鱼鱼乐网络科技有限公司 一种基于元路径相似度的用户推荐方法、系统、服务器及存储介质
CN112738545A (zh) * 2020-12-28 2021-04-30 北京蜜莱坞网络科技有限公司 直播间分享检测方法、装置、电子设备及存储介质
CN114554253B (zh) * 2021-12-27 2024-02-23 北京达佳互联信息技术有限公司 一种资源推荐方法、装置、电子设备和存储介质
CN114650432B (zh) * 2022-04-25 2023-10-17 咪咕视讯科技有限公司 直播视频显示方法、装置、电子设备和介质
CN114969514B (zh) * 2022-05-06 2024-10-15 北京百度网讯科技有限公司 一种直播推荐方法、装置及电子设备

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104053023A (zh) * 2014-06-13 2014-09-17 海信集团有限公司 一种确定视频相似度的方法及装置
CN104598643A (zh) * 2015-02-13 2015-05-06 成都品果科技有限公司 一种物品相似度贡献系数、相似度获取方法及物品推荐方法及其系统
US20160110449A1 (en) * 2012-10-24 2016-04-21 Imagination Technologies Limited Method, System and Device for Connecting Similar Users
CN106294800A (zh) * 2016-08-16 2017-01-04 武汉斗鱼网络科技有限公司 基于加权k近邻评分的直播间推荐方法及系统
CN106560811A (zh) * 2016-09-23 2017-04-12 武汉斗鱼网络科技有限公司 一种基于主播风格的直播间推荐方法及系统

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20160110449A1 (en) * 2012-10-24 2016-04-21 Imagination Technologies Limited Method, System and Device for Connecting Similar Users
CN104053023A (zh) * 2014-06-13 2014-09-17 海信集团有限公司 一种确定视频相似度的方法及装置
CN104598643A (zh) * 2015-02-13 2015-05-06 成都品果科技有限公司 一种物品相似度贡献系数、相似度获取方法及物品推荐方法及其系统
CN106294800A (zh) * 2016-08-16 2017-01-04 武汉斗鱼网络科技有限公司 基于加权k近邻评分的直播间推荐方法及系统
CN106560811A (zh) * 2016-09-23 2017-04-12 武汉斗鱼网络科技有限公司 一种基于主播风格的直播间推荐方法及系统

Also Published As

Publication number Publication date
CN107613395A (zh) 2018-01-19
CN107613395B (zh) 2019-11-15

Similar Documents

Publication Publication Date Title
WO2019041706A1 (zh) 直播房间推荐方法及系统
Turnbull et al. Core domains for clinical research in acute respiratory failure survivors: an international modified Delphi consensus study
Higgins et al. Health economic methods: cost-minimization, cost-effectiveness, cost-utility, and cost-benefit evaluations
US8380784B2 (en) Correlated information recommendation
JP5981541B2 (ja) ユーザ側で開始するソーシャル・ネットワーキング・オブジェクト宣伝
Kleinerman et al. Providing explanations for recommendations in reciprocal environments
Kort-Butler et al. Framing and cultivating the story of crime: The effects of media use, victimization, and social networks on attitudes about crime
US20150213368A1 (en) Information recommendation method, apparatus, and server
US9972028B2 (en) Identifying a social leader
US10210627B1 (en) Image processing system for determining metrics of objects represented therein
WO2015057725A1 (en) Social media platform with gamification of user generated content
MX2014007025A (es) Metodo y sistema para medir y mejorar la eficiencia de campañas de medios sociales.
CN113535991B (zh) 一种多媒体资源推荐方法、装置、电子设备及存储介质
US20160217489A1 (en) System and method for measuring and improving the efficiency of social media campaigns
CN109377288A (zh) 一种广告投放分析方法及装置
KR20200003109A (ko) 샘플 가중치 설정방법 및 장치, 전자 기기
CN105183925A (zh) 内容关联推荐方法及装置
CN109062914A (zh) 用户推荐方法和装置，存储介质和服务器
Ismail et al. Tracking the cyber entrepreneurial intention of private universities students in Malaysia
JP2014200007A (ja) 推薦プログラム、推薦情報受信プログラム、推薦方法、推薦情報受信方法及び推薦装置
JP4447552B2 (ja) 情報提供方法及び装置及びプログラム及びコンピュータ読み取り可能な記録媒体
Egger Identifying key opinion leaders in social networks-an approach to use Instagram data to rate and identify key opinion leader for a specific business field
US20170286975A1 (en) Data Infrastructure and Method for Estimating Influence Spread in Social Networks
US11854369B2 (en) Multi-computer processing system for compliance monitoring and control
JP2016506553A (ja) オンラインビデオ用の自動化サムネイル選択

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: 17923710

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: 17923710

Country of ref document: EP

Kind code of ref document: A1