WO2019071831A1 - 基于路径预测的直播推荐方法、存储介质、设备及系统 - Google Patents

基于路径预测的直播推荐方法、存储介质、设备及系统 Download PDF

Info

Publication number
WO2019071831A1
WO2019071831A1 PCT/CN2017/117383 CN2017117383W WO2019071831A1 WO 2019071831 A1 WO2019071831 A1 WO 2019071831A1 CN 2017117383 W CN2017117383 W CN 2017117383W WO 2019071831 A1 WO2019071831 A1 WO 2019071831A1
Authority
WO
WIPO (PCT)
Prior art keywords
live
room
users
live broadcast
partition
Prior art date
Application number
PCT/CN2017/117383
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 WO2019071831A1 publication Critical patent/WO2019071831A1/zh

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
    • G06F16/70—Information retrieval; Database structures therefor; File system structures therefor of video data
    • G06F16/73—Querying
    • G06F16/735—Filtering based on additional data, e.g. user or group profiles
    • 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
    • G06F16/958—Organisation or management of web site content, e.g. publishing, maintaining pages or automatic linking
    • 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]
    • H04N21/20—Servers specifically adapted for the distribution of content, e.g. VOD servers; Operations thereof
    • H04N21/25—Management operations performed by the server for facilitating the content distribution or administrating data related to end-users or client devices, e.g. end-user or client device authentication, learning user preferences for recommending movies
    • H04N21/251—Learning process for intelligent management, e.g. learning user preferences for recommending movies
    • 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]
    • H04N21/20—Servers specifically adapted for the distribution of content, e.g. VOD servers; Operations thereof
    • H04N21/25—Management operations performed by the server for facilitating the content distribution or administrating data related to end-users or client devices, e.g. end-user or client device authentication, learning user preferences for recommending movies
    • H04N21/258—Client or end-user data management, e.g. managing client capabilities, user preferences or demographics, processing of multiple end-users preferences to derive collaborative data
    • H04N21/25866—Management of end-user data
    • H04N21/25891—Management of end-user data being end-user preferences
    • 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
    • H04N21/47—End-user applications
    • H04N21/482—End-user interface for program selection
    • H04N21/4826—End-user interface for program selection using recommendation lists, e.g. of programs or channels sorted out according to their score

Definitions

  • the present invention relates to the field of big data recommendation technologies, and in particular, to a live recommendation method, a storage medium, a device and a system based on path prediction.
  • the viewing behavior of the live broadcast platform on the live broadcast platform is sequential. Therefore, each user has a live broadcast viewing path, which reflects when the user viewed the live broadcast room.
  • the user can be recommended in real time to predict the next possible live broadcast room. This recommendation can recommend a suitable live room for the user based on the user's behavior in real time.
  • the methods for feedback recommendation based on user behavior in real time are generally as follows:
  • the object of the present invention is to provide a live recommendation method based on path prediction, which has high correlation degree, and comprehensively considers the partition and the transfer event between users in the live broadcast, and the diversity of recommendation results. Ok, the user experience is good.
  • the technical solution adopted by the present invention is: a live recommendation method based on path prediction:
  • S1 record the live broadcast room and the live broadcast room where all users on the live broadcast platform play in the play order; count the transfer events of all users between the live broadcast rooms in a preset time period;
  • the transition probabilities between the live broadcast rooms are arranged in descending order to generate a live broadcast recommendation list
  • S4 Pushing, to the user, a plurality of live broadcast rooms in the recommendation list that are in an open state and have the highest transition probability.
  • step S2 includes:
  • transition order from the live room i to the live room j is k, and the number of effective transitions from the live room i through the k-step transfer order to the live room j is calculated.
  • p ij is the weighted sum of all transfer orders of live room i to live room j
  • k is the transition order from live broadcast i to live broadcast j
  • d is the maximum transition order
  • ⁇ n is the weight coefficient of each transition order
  • the weight coefficients of all transition orders The sum is 1.
  • S ij is the similarity between the two partitions c j and c i , The maximum of the similarity between partition c j and all other partitions;
  • N 11 is the number of users who saw the partition j among the users who saw the partition i
  • N 12 is the partition that did not see the partition among the users who saw the partition i
  • the number of users of j is the number of users who have seen the partition j among the users who did not see the partition i
  • N 22 is the number of users who did not see the partition j among the users who did not see the partition i.
  • the maximum transfer order d is 3.
  • the present invention also discloses a storage medium having stored thereon a computer program that, when executed by a processor, implements the method.
  • the invention also discloses an electronic device comprising a memory and a processor, the memory storing a computer program running on the processor, the method being implemented when the processor executes the computer program.
  • the invention also discloses a live recommendation system based on path prediction, comprising:
  • a transfer event statistic module configured to record, in a play order, a live broadcast room and a live broadcast room where all users on the live broadcast platform are effectively played; and count the transfer events of all users between the live broadcast rooms in a preset time period;
  • a transition probability calculation module configured to calculate a transition probability of the user between the live broadcast rooms according to a transfer event between the live broadcast rooms of each user
  • a recommendation list generating module configured to sort the transition probabilities between the live broadcast rooms according to a high to low to generate a live broadcast recommendation list
  • a recommendation list pushing module is configured to push, to the user, a plurality of live rooms in the recommendation list that are in an on-air state and have the highest transition probability.
  • the transition probability calculation module includes:
  • Effective transfer number calculation unit for calculating the number of effective transfer times from the live room i through the k-step transfer order to the live room j
  • the transition order from the live room i to the live room j be In order to transfer from the live room i through the k-step transfer order to the live room j, the effective transfer times contribution:
  • c j , c i are the partitions to which the live rooms j and i belong
  • S ⁇ c i , c j > are the relative similarities between the two partitions c j and c i ;
  • a transition probability calculation unit for calculating a transition probability from the live broadcast i through the k-step transition order to the live broadcast j
  • p ij is the weighted sum of all transition orders of the live room i to the live room j
  • k is the transition order from live broadcast i to live broadcast j
  • d is the maximum transition order
  • k is the transition order from live broadcast i to live broadcast j
  • d is the maximum transition order
  • k is the maximum transition order
  • ⁇ n is the weight coefficient of each transition order
  • the weight coefficients of all transition orders The sum is 1.
  • N 11 is the number of users who saw the partition j among the users who saw the partition i
  • N 12 is the partition that did not see the partition among the users who saw the partition i
  • the number of users of j is the number of users who have seen the partition j among the users who did not see the partition i
  • N 22 is the number of users who did not see the partition j among the users who did not see the partition i.
  • the maximum transfer order d is 3.
  • the invention calculates the transition probability of the user between the live broadcast rooms according to the transfer event between the live broadcast rooms of each user, and ranks the transition probability between the live broadcast rooms according to the highest to lowest to generate a recommendation list for the live broadcast.
  • the live broadcast room may be selected by the user in the next step, and the recommendation result is highly correlated, and the partitioning and user preferences are comprehensively considered, and the recommendation result is good.
  • FIG. 1 is a schematic flowchart of a live recommendation method based on path prediction according to an embodiment of the present invention
  • FIG. 2 is a block diagram of an electronic device connection in an embodiment of the present invention.
  • an embodiment of the present invention provides a live recommendation method based on path prediction:
  • S1 record the live broadcast room and the direct play of all users on the live broadcast platform in the order of play.
  • S2 Calculate the transition probability of the user between the live broadcast rooms according to the transfer event between each live broadcast room.
  • step S2 includes:
  • transition order from the live room i to the live room j is k, and the number of effective transitions from the live room i through the k-step transfer order to the live room j is calculated.
  • S ij is the similarity between the two partitions c j and c i , The maximum of the similarity between partition c j and all other partitions;
  • N 11 is the number of users who saw the partition j among the users who saw the partition i
  • N 12 is the partition that did not see the partition among the users who saw the partition i
  • the number of users of j is the number of users who have seen the partition j among the users who did not see the partition i
  • N 22 is the number of users who did not see the partition j among the users who did not see the partition i.
  • p ij is the weighted sum of all transfer orders of live room i to live room j
  • k is the transition order from live broadcast i to live broadcast j
  • d is the maximum transition order
  • ⁇ n is the weight coefficient of each transition order
  • the weight coefficients of all transition orders The sum is 1.
  • the transition probabilities between the live broadcast rooms are arranged in descending order to generate a live broadcast recommendation list.
  • the invention calculates the transition probability of the user between the live broadcast rooms according to the transfer event between the live broadcast rooms of each user, and ranks the transition probability between the live broadcast rooms according to the highest to lowest to generate a recommendation list for the live broadcast.
  • the live broadcast room that the user may select in the next step is predicted, and the recommendation result is highly correlated, and the partition and the transfer event between the live broadcast rooms are comprehensively considered, and the recommendation result is good in diversity and the user experience is good. .
  • the embodiment of the invention further discloses a storage medium on which a computer program is stored, and when the computer program is executed by the processor, a live recommendation method based on path prediction is implemented.
  • an embodiment of the present invention further discloses an electronic device, including a memory and a processor.
  • the memory stores a computer program running on the processor, and the processor implements a live prediction recommendation based on path prediction when executing the computer program. method.
  • the embodiment of the invention also discloses a live recommendation system based on path prediction, including:
  • the event statistics module is configured to record the live broadcast room and the live broadcast room where all users on the live broadcast platform are played in the play order; the statistics are in a preset time period. There are transfer events between users in each live broadcast room;
  • a transition probability calculation module configured to calculate a transition probability of the user between the live broadcast rooms according to a transfer event between the live broadcast rooms of each user
  • a recommendation list generating module configured to sort the transition probabilities between the live broadcast rooms according to a high to low to generate a live broadcast recommendation list
  • a recommendation list pushing module is configured to push a plurality of live rooms in the recommendation list that are in an open state and have the highest probability of transition.
  • the transition probability calculation module includes:
  • Effective transfer number calculation unit for calculating the number of effective transfer times from the live room i through the k-step transfer order to the live room j
  • the transition order from the live room i to the live room j be In order to transfer from the live room i through the k-step transfer order to the live room j, the effective transfer times contribution:
  • c j , c i are the partitions to which the live rooms j and i belong
  • S ⁇ c i , c j > are the relative similarities between the two partitions c j and c i ;
  • a transition probability calculation unit for calculating a transition probability from the live broadcast i through the k-step transition order to the live broadcast j
  • p ij is the weighted sum of all transition orders of the live room i to the live room j
  • k is the transition order from live broadcast i to live broadcast j
  • d is the maximum transition order
  • k is the transition order from live broadcast i to live broadcast j
  • d is the maximum transition order
  • k is the maximum transition order
  • ⁇ n is the weight coefficient of each transition order
  • the weight coefficients of all transition orders The sum is 1.
  • S ij is the similarity between the two partitions c j and c i , The maximum of the similarity between partition c j and all other partitions;
  • N 11 is the number of users who saw the partition j among the users who saw the partition i
  • N 12 is the partition that did not see the partition among the users who saw the partition i
  • the number of users of j is the number of users who have seen the partition j among the users who did not see the partition i
  • N 22 is the number of users who did not see the partition j among the users who did not see the partition i.
  • the maximum transfer order d is 3.

Landscapes

  • Engineering & Computer Science (AREA)
  • Databases & Information Systems (AREA)
  • Theoretical Computer Science (AREA)
  • Multimedia (AREA)
  • Signal Processing (AREA)
  • General Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Human Computer Interaction (AREA)
  • Software Systems (AREA)
  • Computer Graphics (AREA)
  • Computing Systems (AREA)
  • Computational Linguistics (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

本发明公开了一种基于路径预测的直播推荐方法、存储介质、设备及系统，涉及大数据推荐技术领域，本发明根据各用户在各直播间之间的转移事件计算用户在各直播间之间的转移概率，将各直播间之间的转移概率按照由高到低排列以生成直播间推荐列表推荐给用户，实现根据用户的直播间观看历史预测用户下一步可能选择的直播间，推荐结果相关度高，且综合考虑分区以及用户在直播间之间的转移事件，推荐结果多样性好，用户体验感好。

Description

基于路径预测的直播推荐方法、存储介质、设备及系统 技术领域
本发明涉及大数据推荐技术领域，具体涉及一种基于路径预测的直播推荐方法、存储介质、设备及系统。
背景技术
用户在直播平台上对直播间的观看行为是存在先后顺序的，因此每个用户都有一条直播间观看路径，该路径反映了用户在什么时候对哪个直播间进行了观看。通过大数据算法，如果我们能够合理地对用户下一步观看路径进行预测，那么可以在用户看完某直播间后向用户实时推荐预测下一个可能看的多个直播间。这种推荐方案可以实时地根据用户的行为为用户推荐合适的直播间。
根据用户行为实时进行反馈推荐的方法一般有如下几种：
(1)在用户观看直播间后推荐与该直播间相同分区的热门直播间。这种推荐方法的缺点是会偏向于推荐那些热门直播间，仅考虑相同分区以及热门程度，推荐结果多样性差，用户体验感差。
(2)在用户发生搜索行为时推荐与搜索词相关联的直播间。这种推荐方法的缺点是推荐的结果集单一，推荐结果相关度低。
发明内容
针对现有技术中存在的缺陷，本发明的目的在于提供一种基于路径预测的直播推荐方法，推荐结果相关度高，且综合考虑分区以及用户在直播间之间的转移事件，推荐结果多样性好，用户体验感好。
为达到以上目的，本发明采取的技术方案是：一种基于路径预测的直播推荐方法：
S1，按播放顺序记录直播平台上所有用户有效播放的直播间及直播间所在的分区；统计在预设的时间周期内所有用户在各直播间之间的转移事件；
S2，根据各用户在各直播间之间的转移事件计算用户在各直播间之间的转移概率；
S3，将各直播间之间的转移概率按照由高到低排列以生成直播间推荐列表；
S4，向用户推送所述推荐列表中处于开播状态且转移概率最高的若干个直播间。
在上述技术方案的基础上，步骤S2具体过程包括：
S201，设从直播间i到直播间j的转移阶数为k，计算从直播间i经过k步转移阶数到直播间j的有效转移次数
Figure PCTCN2017117383-appb-000001
Figure PCTCN2017117383-appb-000002
为从直播间i经过k步转移阶数到直播间j的行为有效转移次数贡献：
Figure PCTCN2017117383-appb-000003
其中，cj,ci分别是直播间j和i所属的分区，S＜ci,cj＞是cj和ci两个分区的相对相似度；
Figure PCTCN2017117383-appb-000004
为从直播间i经过k阶到其他所有直播间的有效转移次数贡献之和；
S202，计算从直播间i经过k步转移阶数到其他所有直播间的有效转换次数之和
Figure PCTCN2017117383-appb-000005
Figure PCTCN2017117383-appb-000006
S203，计算从直播间i经过k步转移阶数到直播间j的转移概率
Figure PCTCN2017117383-appb-000007
Figure PCTCN2017117383-appb-000008
S204，计算直播间i到直播间j的所有转移阶数的总转移概率pij：
pij为直播间i到直播间j的所有转移阶数的加权之和
Figure PCTCN2017117383-appb-000009
其中，k从直播间i到直播间j的转移阶数，d为最大的转移阶数，k∈[1,d]，δn为各转移阶数的权重系数，所有转移阶数的权重系数之和为1。
在上述技术方案的基础上，两个分区的相对相似度S＜ci,cj＞的计算方法为：
Figure PCTCN2017117383-appb-000010
其中，Sij为cj和ci两个分区的相似度，
Figure PCTCN2017117383-appb-000011
为分区cj与其他所有分区的相似度中的最大值；
Sij＝logN11+logN22-logN12-logN21，其中，其中：N11是看了分区i的用户中看了分区j的用户数；N12是看了分区i的用户中没看分区j的用户数；N21是没看分区i的用户中看了分区j的用户数；N22是没看分区i的用户中没看分区j的用户数。
在上述技术方案的基础上，所述最大的转移阶数d为3。
本发明还公开了一种存储介质，该存储介质上存储有计算机程序，所述计算机程序被处理器执行时实现所述的方法。
本发明还公开了一种电子设备，包括存储器和处理器，存储器上储存有在处理器上运行的计算机程序，处理器执行计算机程序时实现所述的方法。
本发明还公开了一种基于路径预测的直播推荐系统，包括：
转移事件统计模块，其用于按播放顺序记录直播平台上所有用户有效播放的直播间及直播间所在的分区；统计在预设的时间周期内所有用户在各直播间之间的转移事件；
转移概率计算模块，其用于根据各用户在各直播间之间的转移事件计算用户在各直播间之间的转移概率；
推荐列表生成模块，其用于将各直播间之间的转移概率按照由高到低排列以生成直播间推荐列表；
推荐列表推送模块，其用于向用户推送所述推荐列表中处于开播状态且转移概率最高的若干个直播间。
在上述技术方案的基础上，所述转移概率计算模块包括：
有效转移次数计算单元，其用于计算从直播间i经过k步转移阶数到直播间j的有效转移次数
Figure PCTCN2017117383-appb-000012
设从直播间i到直播间j的转移阶数为
Figure PCTCN2017117383-appb-000013
为从直播间i经过k步转移阶数到直播间j的行为有效转移次数贡献：
Figure PCTCN2017117383-appb-000014
其中，cj,ci分别是直播间j和i所属的分区，S＜ci,cj＞是cj和ci两个分区的相对相似度；
Figure PCTCN2017117383-appb-000015
为从直播间i经过k阶到其他所有直播间的有效转移次数贡献之和；
有效转换次数之和计算单元，其用于计算从直播间i经过k步转移阶数到其他所有直播间的有效转换次数之和
Figure PCTCN2017117383-appb-000016
转移概率计算单元，其用于计算从直播间i经过k步转移阶数到直播间j的转移概率
Figure PCTCN2017117383-appb-000017
总转移概率计算单元，其用于计算直播间i到直播间j的所有转移阶数的总转移概率pij：pij为直播间i到直播间j的所有转移阶数的加权之和
Figure PCTCN2017117383-appb-000018
其中，k从直播间i到直播间j的转移阶数，d为最大的转移阶数，k∈[1,d]，δn为各转移阶数的权重系数，所有转移阶数的权重系数之和为1。
在上述技术方案的基础上，
Figure PCTCN2017117383-appb-000019
其中，Sij为cj和ci两个分区的相似度，
Figure PCTCN2017117383-appb-000020
为分区cj与其他所有分区的相似度中的最大值；
Sij＝logN11+logN22-logN12-logN21，其中，其中：N11是看了分区i的用户中看了分区j的用户数；N12是看了分区i的用户中没看分区j的用户数；N21是没看分区i的用户中看了分区j的用户数；N22是没看分区i的用户中没看分区j的用户数。
在上述技术方案的基础上，所述最大的转移阶数d为3。
与现有技术相比，本发明的优点在于：
本发明根据各用户在各直播间之间的转移事件计算用户在各直播间之间的转移概率，将各直播间之间的转移概率按照由高到低排列以生成直播间推荐列表推荐给用户，实现根据用户的直播间观看历史预测用户下一步可能选择的直播间，推荐结果相关度高，且综合考虑分区以及用户偏好，推荐结果多样性好。
附图说明
图1为本发明实施例中基于路径预测的直播推荐方法的流程示意图；
图2为本发明实施例中电子设备连接框图。
具体实施方式
以下结合附图及实施例对本发明作进一步详细说明。
参见图1所示，本发明实施例提供一种基于路径预测的直播推荐方法：
S1，按播放顺序记录直播平台上所有用户有效播放的直播间及直 播间所在的分区；只统计用户的有效播放行为播放的直播间；统计在预设的时间周期内所有用户在各直播间之间的转移事件。
S2，根据各用户在各直播间之间的转移事件计算用户在各直播间之间的转移概率。
步骤S2具体过程包括：
S201，设从直播间i到直播间j的转移阶数为k，计算从直播间i经过k步转移阶数到直播间j的有效转移次数
Figure PCTCN2017117383-appb-000021
Figure PCTCN2017117383-appb-000022
为从直播间i经过k步转移阶数到直播间j的行为有效转移次数贡献：
Figure PCTCN2017117383-appb-000023
其中，cj,ci分别是直播间j和i所属的分区，S＜ci,cj＞是cj和ci两个分区的相对相似度；
Figure PCTCN2017117383-appb-000024
为从直播间i经过k阶到其他所有直播间的有效转移次数贡献之和。
两个分区的相对相似度S＜ci,cj＞的计算方法为：
Figure PCTCN2017117383-appb-000025
其中，Sij为cj和ci两个分区的相似度，
Figure PCTCN2017117383-appb-000026
为分区cj与其他所有分区的相似度中的最大值；
Sij＝logN11+logN22-logN12-logN21，其中，其中：N11是看了分区i的用户中看了分区j的用户数；N12是看了分区i的用户中没看分区j的用户数；N21是没看分区i的用户中看了分区j的用户数；N22是没看分区i的用户中没看分区j的用户数。
S202，计算从直播间i经过k步转移阶数到其他所有直播间的有效转换次数之和
Figure PCTCN2017117383-appb-000027
Figure PCTCN2017117383-appb-000028
S203，计算从直播间i经过k步转移阶数到直播间j的转移概率
Figure PCTCN2017117383-appb-000029
Figure PCTCN2017117383-appb-000030
S204，计算直播间i到直播间j的所有转移阶数的总转移概率pij：
pij为直播间i到直播间j的所有转移阶数的加权之和
Figure PCTCN2017117383-appb-000031
其中，k从直播间i到直播间j的转移阶数，d为最大的转移阶数，k∈[1,d]，δn为各转移阶数的权重系数，所有转移阶数的权重系数之和为1。采用最大的转移阶数为3时，可得出理想的推荐结果。
S3，将各直播间之间的转移概率按照由高到低排列以生成直播间推荐列表。
S4，向用户推送推荐列表中处于开播状态且转移概率最高的若干个直播间。
本发明根据各用户在各直播间之间的转移事件计算用户在各直播间之间的转移概率，将各直播间之间的转移概率按照由高到低排列以生成直播间推荐列表推荐给用户，实现根据用户的直播间观看历史预测用户下一步可能选择的直播间，推荐结果相关度高，且综合考虑分区以及用户在直播间之间的转移事件，推荐结果多样性好，用户体验感好。
本发明实施例还公开了一种存储介质，该存储介质上存储有计算机程序，计算机程序被处理器执行时实现基于路径预测的直播推荐方法。
参见图2所示，本发明实施例还公开了一种电子设备，包括存储器和处理器，存储器上储存有在处理器上运行的计算机程序，处理器执行计算机程序时实现基于路径预测的直播推荐方法。
本发明实施例还公开了一种基于路径预测的直播推荐系统，包括：
转移事件统计模块，其用于按播放顺序记录直播平台上所有用户有效播放的直播间及直播间所在的分区；统计在预设的时间周期内所 有用户在各直播间之间的转移事件；
转移概率计算模块，其用于根据各用户在各直播间之间的转移事件计算用户在各直播间之间的转移概率；
推荐列表生成模块，其用于将各直播间之间的转移概率按照由高到低排列以生成直播间推荐列表；
推荐列表推送模块，其用于向用户推送推荐列表中处于开播状态且转移概率最高的若干个直播间。
转移概率计算模块包括：
有效转移次数计算单元，其用于计算从直播间i经过k步转移阶数到直播间j的有效转移次数
Figure PCTCN2017117383-appb-000032
设从直播间i到直播间j的转移阶数为
Figure PCTCN2017117383-appb-000033
为从直播间i经过k步转移阶数到直播间j的行为有效转移次数贡献：
Figure PCTCN2017117383-appb-000034
其中，cj,ci分别是直播间j和i所属的分区，S＜ci,cj＞是cj和ci两个分区的相对相似度；
Figure PCTCN2017117383-appb-000035
为从直播间i经过k阶到其他所有直播间的有效转移次数贡献之和；
有效转换次数之和计算单元，其用于计算从直播间i经过k步转移阶数到其他所有直播间的有效转换次数之和
Figure PCTCN2017117383-appb-000036
转移概率计算单元，其用于计算从直播间i经过k步转移阶数到直播间j的转移概率
Figure PCTCN2017117383-appb-000037
总转移概率计算单元，其用于计算直播间i到直播间j的所有转移阶数的总转移概率pij：pij为直播间i到直播间j的所有转移阶数的加权之和
Figure PCTCN2017117383-appb-000038
其中，k从直播间i到直播间j的转移阶数，d为最大的转移阶数，k∈[1,d]，δn为各转移阶数的权重系数，所有转移 阶数的权重系数之和为1。
Figure PCTCN2017117383-appb-000039
其中，Sij为cj和ci两个分区的相似度，
Figure PCTCN2017117383-appb-000040
为分区cj与其他所有分区的相似度中的最大值；
Sij＝logN11+logN22-logN12-logN21，其中，其中：N11是看了分区i的用户中看了分区j的用户数；N12是看了分区i的用户中没看分区j的用户数；N21是没看分区i的用户中看了分区j的用户数；N22是没看分区i的用户中没看分区j的用户数。最大的转移阶数d为3。
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。

Claims (10)

  1. 一种基于路径预测的直播推荐方法，其特征在于：
    S1，按播放顺序记录直播平台上所有用户播放的直播间及直播间所在的分区；统计在预设的时间周期内所有用户在各直播间之间的转移事件；
    S2，根据各用户在各直播间之间的转移事件计算用户在各直播间之间的转移概率；
    S3，将各直播间之间的转移概率按照由高到低排列以生成直播间推荐列表；
    S4，向用户推送所述推荐列表中处于开播状态且转移概率最高的若干个直播间。
  2. 如权利要求1所述的一种基于路径预测的直播推荐方法，其特征在于：
    步骤S2具体过程包括：
    S201，设从直播间i到直播间j的转移阶数为k，计算从直播间i经过k步转移阶数到直播间j的有效转移次数
    Figure PCTCN2017117383-appb-100001
    Figure PCTCN2017117383-appb-100002
    为从直播间i经过k步转移阶数到直播间j的行为有效转移次数贡献：
    Figure PCTCN2017117383-appb-100003
    其中，cj,ci分别是直播间j和i所属的分区，S＜ci,cj＞是cj和ci两个分区的相对相似度；
    Figure PCTCN2017117383-appb-100004
    为从直播间i经过k阶到其他所有直播间的有效转移次数贡献之和；
    S202，计算从直播间i经过k步转移阶数到其他所有直播间的有效转换次数之和
    Figure PCTCN2017117383-appb-100005
    Figure PCTCN2017117383-appb-100006
    S203，计算从直播间i经过k步转移阶数到直播间j的转移概率
    Figure PCTCN2017117383-appb-100007
    Figure PCTCN2017117383-appb-100008
    S204，计算直播间i到直播间j的所有转移阶数的总转移概率pij：
    pij为直播间i到直播间j的所有转移阶数的加权之和
    Figure PCTCN2017117383-appb-100009
    其中，k从直播间i到直播间j的转移阶数，d为最大的转移阶数，k∈[1,d]，δn为各转移阶数的权重系数，所有转移阶数的权重系数之和为1。
  3. 如权利要求1所述的一种基于路径预测的直播推荐方法，其特征在于：两个分区的相对相似度S＜ci,cj＞的计算方法为：
    Figure PCTCN2017117383-appb-100010
    其中，Sij为cj和ci两个分区的相似度，
    Figure PCTCN2017117383-appb-100011
    为分区cj与其他所有分区的相似度中的最大值；
    Sij＝logN11+logN22-logN12-logN21，其中，其中：N11是看了分区i的用户中看了分区j的用户数；N12是看了分区i的用户中没看分区j的用户数；N21是没看分区i的用户中看了分区j的用户数；N22是没看分区i的用户中没看分区j的用户数。
  4. 如权利要求1所述的一种基于路径预测的直播推荐方法，其特征在于：所述最大的转移阶数d为3。
  5. 一种存储介质，该存储介质上存储有计算机程序，其特征在于：所述计算机程序被处理器执行时实现权利要求1至4任一项所述的方法。
  6. 一种电子设备，包括存储器和处理器，存储器上储存有在处理器上运行的计算机程序，其特征在于：处理器执行计算机程序时实现权利要求1至4任一项所述的方法。
  7. 一种基于路径预测的直播推荐系统，其特征在于，包括：
    转移事件统计模块，其用于按播放顺序记录直播平台上所有用户有效播放的直播间及直播间所在的分区；统计在预设的时间周期内所有用户在各直播间之间的转移事件；
    转移概率计算模块，其用于根据各用户在各直播间之间的转移事件计算用户在各直播间之间的转移概率；
    推荐列表生成模块，其用于将各直播间之间的转移概率按照由高到低排列以生成直播间推荐列表；
    推荐列表推送模块，其用于向用户推送所述推荐列表中处于开播状态且转移概率最高的若干个直播间。
  8. 如权利要求7所述的一种基于路径预测的直播推荐系统，其特征在于：
    所述转移概率计算模块包括：
    有效转移次数计算单元，其用于计算从直播间i经过k步转移阶数到直播间j的有效转移次数
    Figure PCTCN2017117383-appb-100012
    设从直播间i到直播间j的转移阶数为k；
    Figure PCTCN2017117383-appb-100013
    为从直播间i经过k步转移阶数到直播间j的行为有效转移次数贡献：
    Figure PCTCN2017117383-appb-100014
    其中，cj,ci分别是直播间j和i所属的分区，S＜ci,cj＞是cj和ci两个分区的相对相似度；
    Figure PCTCN2017117383-appb-100015
    为从直播间i经过k阶到其他所有直播间的有效转移次数贡献之和；
    有效转换次数之和计算单元，其用于计算从直播间i经过k步转移阶数到其他所有直播间的有效转换次数之和
    Figure PCTCN2017117383-appb-100016
    转移概率计算单元，其用于计算从直播间i经过k步转移阶数到直播间j的转移概率
    Figure PCTCN2017117383-appb-100017
    总转移概率计算单元，其用于计算直播间i到直播间j的所有转移阶数的总转移概率pij：pij为直播间i到直播间j的所有转移阶数的加权之和
    Figure PCTCN2017117383-appb-100018
    其中，k从直播间i到直播间j的转移阶数，d为最大的转移阶数，k∈[1,d]，δn为各转移阶数的权重系数，所有转移阶数的权重系数之和为1。
  9. 如权利要求7所述的一种基于路径预测的直播推荐系统，其特征在于：
    Figure PCTCN2017117383-appb-100019
    其中，Sij为cj和ci两个分区的相似度，
    Figure PCTCN2017117383-appb-100020
    为分区cj与其他所有分区的相似度中的最大值；
    Sij＝logN11+logN22-logN12-logN21，其中，其中：N11是看了分区i的用户中看了分区j的用户数；N12是看了分区i的用户中没看分区j的用户数；N21是没看分区i的用户中看了分区j的用户数；N22是没看分区i的用户中没看分区j的用户数。
  10. 如权利要求7所述的一种基于路径预测的直播推荐系统，其特征在于：所述最大的转移阶数d为3。
PCT/CN2017/117383 2017-10-10 2017-12-20 基于路径预测的直播推荐方法、存储介质、设备及系统 WO2019071831A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201710932686.1 2017-10-10
CN201710932686.1A CN107835441B (zh) 2017-10-10 2017-10-10 基于路径预测的直播推荐方法、存储介质、设备及系统

Publications (1)

Publication Number Publication Date
WO2019071831A1 true WO2019071831A1 (zh) 2019-04-18

Family

ID=61647768

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2017/117383 WO2019071831A1 (zh) 2017-10-10 2017-12-20 基于路径预测的直播推荐方法、存储介质、设备及系统

Country Status (2)

Country Link
CN (1) CN107835441B (zh)
WO (1) WO2019071831A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110717415A (zh) * 2019-09-24 2020-01-21 上海数创医疗科技有限公司 基于特征选取的st段分类卷积神经网络及其使用方法

Families Citing this family (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108536814B (zh) * 2018-04-04 2022-06-21 武汉斗鱼网络科技有限公司 直播间推荐方法、计算机可读存储介质及电子设备
CN109151488B (zh) * 2018-07-06 2021-07-23 武汉斗鱼网络科技有限公司 根据用户行为实时推荐直播间的方法及系统
CN111385657B (zh) * 2018-12-28 2023-02-07 广州市百果园信息技术有限公司 视频推荐方法、装置及存储介质、计算机设备
CN109951725B (zh) * 2019-03-07 2021-06-15 武汉斗鱼鱼乐网络科技有限公司 一种直播间的推荐方法以及相关设备
CN112954460B (zh) * 2021-02-01 2024-01-05 百果园技术(新加坡)有限公司 一种直播互动的方法、装置、服务器和存储介质
CN113315992B (zh) * 2021-07-30 2021-11-09 武汉斗鱼鱼乐网络科技有限公司 用于提升观看时长的直播间推荐方法、装置、介质及设备
CN114969514A (zh) * 2022-05-06 2022-08-30 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 (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2008121737A1 (en) * 2007-03-30 2008-10-09 Amazon Technologies, Inc. Service for providing item recommendations
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
CN106604051A (zh) * 2016-12-20 2017-04-26 广州华多网络科技有限公司 直播频道推荐方法及装置
CN106658096A (zh) * 2016-11-17 2017-05-10 百度在线网络技术（北京）有限公司 推送直播节目的方法和装置
CN106851343A (zh) * 2017-01-23 2017-06-13 百度在线网络技术（北京）有限公司 用于视频直播的方法和装置
CN107205178A (zh) * 2017-04-25 2017-09-26 北京潘达互娱科技有限公司 直播间推荐方法及装置

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105279289B (zh) * 2015-12-04 2019-03-22 中国传媒大学 基于指数衰减窗口的个性化音乐推荐排序方法
CN106202430A (zh) * 2016-07-13 2016-12-07 武汉斗鱼网络科技有限公司 基于关联规则的直播平台用户兴趣度挖掘系统及挖掘方法
CN106791966B (zh) * 2016-12-28 2019-11-15 武汉斗鱼网络科技有限公司 一种基于改进型关联规则的直播间推荐方法及系统

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2008121737A1 (en) * 2007-03-30 2008-10-09 Amazon Technologies, Inc. Service for providing item recommendations
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
CN106658096A (zh) * 2016-11-17 2017-05-10 百度在线网络技术（北京）有限公司 推送直播节目的方法和装置
CN106604051A (zh) * 2016-12-20 2017-04-26 广州华多网络科技有限公司 直播频道推荐方法及装置
CN106851343A (zh) * 2017-01-23 2017-06-13 百度在线网络技术（北京）有限公司 用于视频直播的方法和装置
CN107205178A (zh) * 2017-04-25 2017-09-26 北京潘达互娱科技有限公司 直播间推荐方法及装置

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110717415A (zh) * 2019-09-24 2020-01-21 上海数创医疗科技有限公司 基于特征选取的st段分类卷积神经网络及其使用方法
CN110717415B (zh) * 2019-09-24 2020-12-04 上海数创医疗科技有限公司 基于特征选取的st段分类卷积神经网络及其使用方法

Also Published As

Publication number Publication date
CN107835441B (zh) 2020-01-03
CN107835441A (zh) 2018-03-23

Similar Documents

Publication Publication Date Title
WO2019071831A1 (zh) 基于路径预测的直播推荐方法、存储介质、设备及系统
US10397359B2 (en) Streaming media cache for media streaming service
KR101941757B1 (ko) 콘텐츠 자동 추천
US9798980B2 (en) Method for inferring latent user interests based on image metadata
US9740775B2 (en) Video retrieval based on optimized selected fingerprints
McInerney et al. Counterfactual evaluation of slate recommendations with sequential reward interactions
US8799973B2 (en) Methods and apparatus for selecting and pushing customized electronic media content
US20150143394A1 (en) Content presentation method, content presentation device, and program
US20120303623A1 (en) System for incrementally clustering news stories
US9582767B2 (en) Media recommendation using internet media stream modeling
CN107766360B (zh) 一种视频热度预测方法和装置
CA2771379C (en) Estimating and displaying social interest in time-based media
KR101468201B1 (ko) 문서들로부터 토픽들의 병렬 생성
US20140044407A1 (en) Segmenting video based on timestamps in comments
WO2015032353A1 (zh) 视频推荐方法及装置
US20130325942A1 (en) Recommender system for content delivery networks
WO2021135701A1 (zh) 一种信息推荐的方法及装置、电子设备、存储介质
CN108536814B (zh) 直播间推荐方法、计算机可读存储介质及电子设备
Zhou et al. Online social media recommendation over streams
Abdelkrim et al. A hybrid regression model for video popularity-based cache replacement in content delivery networks
CN111385657B (zh) 视频推荐方法、装置及存储介质、计算机设备
Liu et al. Determinantal point process likelihoods for sequential recommendation
CN103607606A (zh) 一种基于词网络的视频播放量预估方法及装置
JP2020501396A (ja) コンテンツストリームにおける中断期間の予測
RU2617391C2 (ru) Способ и устройство для генерирования отсортированного списка элементов

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: 17928410

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: 17928410

Country of ref document: EP

Kind code of ref document: A1