WO2021197001A1 - 一种融合社交关系与自私偏好顺序的群组划分方法 - Google Patents

一种融合社交关系与自私偏好顺序的群组划分方法 Download PDF

Info

Publication number
WO2021197001A1
WO2021197001A1 PCT/CN2021/079866 CN2021079866W WO2021197001A1 WO 2021197001 A1 WO2021197001 A1 WO 2021197001A1 CN 2021079866 W CN2021079866 W CN 2021079866W WO 2021197001 A1 WO2021197001 A1 WO 2021197001A1
Authority
WO
WIPO (PCT)
Prior art keywords
group
user
selfish
preference
expressed
Prior art date
Application number
PCT/CN2021/079866
Other languages
English (en)
French (fr)
Inventor
王海艳
王晨一
王周生
杨一铖
Original Assignee
南京邮电大学
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 南京邮电大学 filed Critical 南京邮电大学
Priority to US17/795,453 priority Critical patent/US20230091117A1/en
Publication of WO2021197001A1 publication Critical patent/WO2021197001A1/zh

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90Details of database functions independent of the retrieved data types
    • G06F16/95Retrieval from the web
    • G06F16/953Querying, e.g. by the use of web search engines
    • G06F16/9536Search customisation based on social or collaborative filtering
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06QINFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00Administration; Management
    • G06Q10/04Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06QINFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/01Social networking

Definitions

  • the invention belongs to the cross field of information services and distributed computing, and specifically relates to a group division technology in the field of group recommendation.
  • Group recommendation aims to provide recommendations for a group of users, and group division is also very important as the basis for group recommendations.
  • group division is also very important as the basis for group recommendations.
  • most methods often focus on the same user interest preferences, while ignoring the influence of users' social relationships on the recommendation results, and even failing to consider the positive effects of user selfish behavior on the division effect. All of these will cause serious damage to the recommendation results if there are extreme users in the group, and will affect the satisfaction of the recommendation.
  • the present invention proposes a group division method that integrates social relations and selfish preference order to stably divide groups to eliminate extreme users to improve recommendation satisfaction.
  • a group division method integrating social relationships and selfish preference order including the following steps:
  • Step 1 According to the user's social value, simulate the choice of each user to gather as a group to share the cost to form a preliminary group;
  • Step 2 Formulate merging and splitting rules based on the order of selfish preference, and merge or split the group obtained in step 1 based on the merging and splitting rules.
  • M ⁇ S, v, F, A, P ⁇
  • S represents the decision-making user Set
  • v represents the utility function
  • F represents the decision function for deciding users to join or leave the group
  • a n represents group n
  • P represents group Group structure
  • P ⁇ CO 1 , CO 2 ,..., CO m ⁇
  • CO m represents the structural feature of group m
  • the utility function is defined as the cost of the user's current group.
  • the utility function is expressed as:
  • a -n n represents a leaving group after the user group a n.
  • a n indicates the user is currently groups, Order of preference based on a user indicates a future opt selfish group, a n groups and groups r n (a n) in the group that the user gains of a n, Indicates that the user is in the group
  • the earnings, r i (a n) represents the group of a n overall revenue;
  • the merging rule is expressed as:
  • the split rule is expressed as:
  • the step of obtaining the Nash equilibrium point in the step 2 includes:
  • the group formation process is an exact latent game, and the optimal response algorithm is used to converge the latent function in the latent game to obtain the Nash equilibrium point;
  • the latent function of the latent game is the sum of the social values of all users, expressed as:
  • a group division system that integrates social relations and selfish preference order includes:
  • the preliminary group building module is used to simulate the selection of groups gathered by users to share costs according to the social value of users to form preliminary groups;
  • the stable group building module is used to formulate merging and splitting rules based on the order of selfish preference, and to merge or split the group obtained in step 1 based on the merging and splitting rules.
  • the Nash equilibrium point is reached, the group merges or After the division is over, a stable group is obtained;
  • M ⁇ S, v, F, A, P ⁇
  • S represents the decision-making user Set
  • v represents the utility function
  • F represents the decision function for deciding users to join or leave the group
  • a n represents group n
  • P represents group Group structure
  • P ⁇ CO 1 , CO 2 ,..., CO m ⁇
  • CO m represents the structural feature of group m
  • the utility function is defined as the cost of the user's current group.
  • the utility function is expressed as:
  • a -n n represents a leaving group after the user group a n.
  • a n indicates the user is currently groups, Order of preference based on a user indicates a future opt selfish group, a n groups and groups r n (a n) in the group that the user gains of a n, Indicates that the user is in the group
  • the earnings, r i (a n) represents the group of a n overall revenue;
  • the merging rule is expressed as:
  • the split rule is expressed as:
  • the step of obtaining the Nash equilibrium point includes:
  • the group formation process is an exact latent game, and the optimal response algorithm is used to converge the latent function in the latent game to obtain the Nash equilibrium point;
  • the latent function of the latent game is the sum of the social values of all users, expressed as:
  • the present invention has the following advantages:
  • the division method of the present invention can significantly improve group members' satisfaction with group division, and avoid the occurrence of extreme users in the group;
  • the method of the present invention generates a game between group members. Compared with the traditional group division method, the user’s selfishness is considered.
  • the group member selection process is simulated through the selfish preference sequence, and the Nash equilibrium point is found to make Stable group division, laying a solid foundation for the effect of group recommendation;
  • the present invention considers the social relationship between group members, and solves the problem of the interests of the members well, so that the group members can maximize their interests when making choices.
  • Figure 1 is a work flow chart of the division method that integrates social relationships and selfish preference order.
  • Step 1 According to the user's social value, simulate the user's choice of gathering as a group in order to share the cost to form a preliminary group, where the social value is reflected in cost.
  • D total [d 1 , d 2 , d 3 ,..., d D max ]
  • M is the group
  • the social value of user n is used to reflect the social relationship between the user and the group, and the social value of user n is expressed as:
  • N where the cost of the user group to give a n may be calculated based on the formula:
  • Step 2 Members will choose a group with higher returns.
  • the order of selfish preference allows group members to selfishly choose the group that maximizes their interests in order to seek higher interests. Therefore, the order of selfish preference is used to formulate game rules to simulate groups Selection of members, specific:
  • user n For user n, user n’s current group a n, and future groups chosen to join based on the order of selfish preference Group and a n group
  • R & lt n (a n) n in the group that the user proceeds in a n Indicates that user n is in the group
  • the earnings, r i (a n) represents the group of a n overall revenue.
  • a group a group a n if n when the user selected to join the group in order of preference based on selfish , You can reduce the group
  • the overall cost and cost of other users in the group does not increase a n, a n and may group the group Merge into a new group, define the above process as a merge rule, and express it as follows:
  • Step 3 Review the formation process of the group from the user's perspective.
  • the formation process of the group can be simulated as an exact potential game, and the potential game has at least one Nash equilibrium point.
  • the division process of step 2 is finite Yes, when the Nash equilibrium point is reached, the group members have no incentive to leave the group to achieve a stable division effect. By looking for the Nash equilibrium point, the effect of stable division of the group can be achieved.
  • step 2 is an accurate potential game with potential functions, which is expressed as:
  • the potential function of the present invention is the sum of the social values of all users, expressed as follows:
  • a -n represents the group after user n leaves the group a n, Indicates that user n has not joined the group
  • ⁇ i represents the total number of the group of a n d ni in demand;
  • ⁇ 'i represents a group of ni the total number of demand d.
  • N denotes the set of users
  • the user set is divided into four parts: user n, a n groups referred to in the rest of the users J an, added group
  • the remaining users in are recorded as And the remaining users.
  • the optimal response algorithm in the latent game always converges to the Nash equilibrium. According to the Nash equilibrium theorem:
  • NA (NA 1 ,..., NA
  • This embodiment also provides a group division system that integrates social relations and selfish preference order, and the group division system that integrates social relations and selfish preference order includes:
  • the preliminary group building module is used to simulate the selection of groups gathered by users to share costs according to the user's social value to form a preliminary group.
  • the stable group building module is used to formulate merging and splitting rules based on the order of selfish preference, and to merge or split the group obtained in step 1 based on the merging and splitting rules.
  • the Nash equilibrium point is reached, the group merges or The division is over, and a stable group is obtained.
  • M ⁇ S, v, F, A, P ⁇
  • S represents the decision-making user Set
  • v represents the utility function
  • F represents the decision function for deciding users to join or leave the group
  • a n represents group n
  • P represents group Group structure
  • P ⁇ CO 1 , CO 2 ,..., CO m ⁇
  • CO m represents the structural feature of group m
  • the utility function is defined as the cost of the user's current group.
  • the utility function is expressed as:
  • a -n n represents a leaving group after the user group a n.
  • the social value of user n is expressed as:
  • the selfish preference is expressed as:
  • a n indicates the user is currently groups, Order of preference based on a user indicates a future opt selfish group, a n groups and groups r n (a n) in the group that the user gains of a n, Indicates that the user is in the group
  • the earnings, r i (a n) represents the group of a n overall revenue;
  • the merging rule is expressed as:
  • the split rule is expressed as:
  • the step of obtaining the Nash equilibrium point includes:
  • the group formation process is an exact latent game, and the optimal response algorithm is used to converge the latent function in the latent game to obtain the Nash equilibrium point;
  • the latent function of the latent game is the sum of the social values of all users, expressed as:

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Strategic Management (AREA)
  • Economics (AREA)
  • Human Resources & Organizations (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Physics & Mathematics (AREA)
  • General Business, Economics & Management (AREA)
  • Tourism & Hospitality (AREA)
  • Marketing (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Computing Systems (AREA)
  • Primary Health Care (AREA)
  • Databases & Information Systems (AREA)
  • Development Economics (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Game Theory and Decision Science (AREA)
  • Data Mining & Analysis (AREA)
  • General Engineering & Computer Science (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

一种融合社交关系与自私偏好顺序的群组划分方法,包括以下步骤:步骤1:根据用户的社交价值模拟各用户为了分担成本而聚集为群组的选择,形成初步群组;步骤2:通过自私偏好顺序制定合并和拆分规则,并基于合并和拆分规则对步骤1得到的群组进行合并或拆分,当达到纳什均衡点时,群组合并或划分结束,得到稳定群组。该方法提出了在完全信息静态的博弈场景下融合社交关系与自私偏好顺序并以此寻找纳什均衡点,来提高群组划分的稳定性并剔除极端成员,解决了群组中极端用户的问题,提高了群组成员对群组推荐结果的满意度。

Description

一种融合社交关系与自私偏好顺序的群组划分方法
本申请要求于2020年04月03日提交中国专利局、申请号为202010259534.1、发明名称为“一种融合社交关系与自私偏好顺序的群组划分方法”的中国专利申请的优先权,其全部内容通过引用结合在本申请中。
技术领域
本发明属于信息服务和分布式计算的交叉领域,具体涉及一种在群组推荐领域中的群组划分技术。
背景技术
传统的推荐方法更多的是向个人进行推荐,然而随着经济的发展,互联网技术的不断革新,现实生活中越来越多的活动是由群组来完成,所以向群组成员推荐都能接受的项目变得至关重要。
群组推荐旨在为一组用户提供推荐,而群组划分作为群组推荐的基础同样非常重要。但在已有的群组划分研究中,大多数方法往往关注的只是用户兴趣偏好的相同,而忽略了用户的社交关系对推荐结果的影响,更加没有考虑到用户的自私行为对划分效果的积极影响,这些都使得若群组中存在极端用户则会对推荐结果造成严重损害,都会影响推荐的满意度。
发明内容
本发明的目:针对现有技术的不足,本发明提出一种融合社交关系与自私偏好顺序的群组划分方法,来对群组进行稳定的划分去除极端用户以此来提高推荐满意度。
技术方案:一种融合社交关系与自私偏好顺序的群组划分方法,包括以下步骤:
步骤1:根据用户的社交价值模拟各用户为了分担成本而聚集为群组的选择,形成初步群组;
步骤2:通过自私偏好顺序制定合并和拆分规则,并基于合并和拆分规则对步骤1得到的群组进行合并或拆分,当达到纳什均衡点时,群组合并或划分 结束,得到稳定群组;
其中,所述的根据用户的社交价值模拟各用户为了分担成本而聚集为群组的选择,将该过程表示为:M={S,v,F,A,P},其中,S表示决策用户集,v表示效用函数,F表示决定用户加入或退出群组的决策函数,A={a 1,a 2,…,a N}表示总用户决策集,a n表示群组n,P表示群组结构P={CO 1,CO 2,…,CO m};CO m表示群组m的结构特征,所述效用函数定义为用户所在其当前群组的成本。
进一步的,所述成员所在群组a n的成本表示为:
Figure PCTCN2021079866-appb-000001
式中,l n表示为用户n自身的喜好集合l n=[d n1,d n2,d n3,...,d nln],d nj∈D total,1≤j≤l n;D total=[d 1,d 2,d 3,...,d Dmax]表示为群组a n所有成员的喜好集合;β i表示群组a n中需求d ni的总数量;J n表示用户n的邻居集,a n表示用户n所在群组,a Jn表示用户n的邻居集所在群组;a 0表示偏好成本,l c表示与群组a n内其他成员相同的偏好,L max表示群组a n中的所有偏好;
所述效用函数表示为:
u n(a n,a -n)=r n(a n,a Jn)     (3)
式中,a -n表示用户n离开群组a n后的群组。
进一步的,所述用户n的社交价值表示为:
Figure PCTCN2021079866-appb-000002
进一步的,所述自私偏好表示为:
Figure PCTCN2021079866-appb-000003
式中,a n表示用户当前所在群组,
Figure PCTCN2021079866-appb-000004
表示用户基于自私偏好顺序将来选择加入的群组,群组a n和群组
Figure PCTCN2021079866-appb-000005
r n(a n)表示用户在群组a n中的收益,
Figure PCTCN2021079866-appb-000006
表示用户在群组
Figure PCTCN2021079866-appb-000007
中的收益,r i(a n)表示群组a n的整体收益;
所述合并规则表示为:
Figure PCTCN2021079866-appb-000008
所述拆分规则表示为:
Figure PCTCN2021079866-appb-000009
进一步的,所述步骤2中获得纳什均衡点的步骤包括:
群组形成过程为确切的潜在博弈,采用最佳响应算法使潜在博弈中的潜在函数收敛,得到纳什均衡点;
所述潜在博弈的潜在函数为所有用户的社会价值的总和,表示为:
Figure PCTCN2021079866-appb-000010
一种融合社交关系与自私偏好顺序的群组划分系统,所述融合社交关系与自私偏好顺序的群组划分系统包括:
初步群组构建模块,用于根据用户的社交价值模拟各用户为了分担成本而聚集为群组的选择,形成初步群组;
稳定群组构建模块,用于通过自私偏好顺序制定合并和拆分规则,并基于合并和拆分规则对步骤1得到的群组进行合并或拆分,当达到纳什均衡点时,群组合并或划分结束,得到稳定群组;
其中,所述的根据用户的社交价值模拟各用户为了分担成本而聚集为群组的选择,将该过程表示为:M={S,v,F,A,P},其中,S表示决策用户集,v表示效用函数,F表示决定用户加入或退出群组的决策函数,A={a 1,a 2,…,a N}表示总用户决策集,a n表示群组n,P表示群组结构P={CO 1,CO 2,…,CO m};CO m表示群组m的结构特征,所述效用函数定义为用户所在其当前群组的成本。
进一步的,所述成员所在群组a n的成本表示为:
Figure PCTCN2021079866-appb-000011
式中,l n表示为用户n自身的喜好集合l n=[d n1,d n2,d n3,...,d nln],d nj∈D total,1≤j≤l n;D total=[d 1,d 2,d 3,...,d Dmax]表示为群组a n所有成员的喜好集合;β i表示群组a n中需求d ni的总数量;J n表示用户n的邻居集,a n表示用户n所在群组,a Jn表示用户n的邻居集所在群组;a 0表示偏好成本,l c表示与群组a n内其他成员相同的偏好,L max表示群组a n中的所有偏好;
所述效用函数表示为:
u n(a n,a -n)=r n(a n,a Jn)
式中,a -n表示用户n离开群组a n后的群组。
进一步的,所述用户n的社交价值表示为:
Figure PCTCN2021079866-appb-000012
进一步的,所述自私偏好表示为:
Figure PCTCN2021079866-appb-000013
式中,a n表示用户当前所在群组,
Figure PCTCN2021079866-appb-000014
表示用户基于自私偏好顺序将来选择加入的群组,群组a n和群组
Figure PCTCN2021079866-appb-000015
r n(a n)表示用户在群组a n中的收益,
Figure PCTCN2021079866-appb-000016
表示用户在群组
Figure PCTCN2021079866-appb-000017
中的收益,r i(a n)表示群组a n的整体收益;
所述合并规则表示为:
Figure PCTCN2021079866-appb-000018
所述拆分规则表示为:
Figure PCTCN2021079866-appb-000019
进一步的,所述纳什均衡点的获得步骤包括:
群组形成过程为确切的潜在博弈,采用最佳响应算法使潜在博弈中的潜在函数收敛,得到纳什均衡点;
所述潜在博弈的潜在函数为所有用户的社会价值的总和,表示为:
Figure PCTCN2021079866-appb-000020
有益效果:本发明具有以下优点:
1、本发明的划分方法可以显著提高群组成员对群组划分的满意度,避免了群组中出现极端用户的情况;
2、本发明的方法在群组成员之间产生了博弈,与传统的群组划分方法相比考虑了用户的自私性,通过自私偏好顺序模拟群组成员选择的过程,并找到纳什均衡点使得群组稳定划分,为群组推荐的效果打下坚实的基础;
3、本发明考虑了群组成员之间的社交关系,很好的解决了成员的利益问题,使得群组成员在做出选择时,能够达到利益的最大化。
附图说明
图1是融合社交关系与自私偏好顺序的划分方法的工作流程图。
具体实施方式
现结合附图和实施例进一步阐述本发明的技术方案。
本实施例的一种融合社交关系与自私偏好顺序的群组划分方法,具体包括以下步骤:
步骤1:根据用户的社交价值模拟用户为了分担成本而聚集为群组的选择,形成初步群组,此处的社交价值是以成本体现。具体的:
根据其成本需求来考虑用户的社交价值,当群组中的成员需要的东西更加相似则他们付出的成本就越少,这样群组的收益就更高;根据用户之间的社交关系可以帮助他们降低成本,若两个用户喜好集合交叉越多,它们的关系就越紧密,两个用户关系密切他们就可以组成一个群组,这样他们对推荐内容付出的成本就越少。
将群组内所有用户的喜好集合记为D total=[d 1,d 2,d 3,...,d D max],用户n自 身的喜好集合表示为l n=[d n1,d n2,d n3,...,d nl n],d nj∈D total,1≤j≤l n;设定用户n选择群组a n=m,1≤m≤M,其中,M是群组数,用户n的邻居集被定义为群组中的其他用户,a Jn表示用户n的邻居集所在群组,J n={i:a i=a n,i∈M\n}。
设定群组中的用户总数量为CO m。根据用户n自身的喜好集合向量,计算用户n与各群组之间具有相同数据的数量并将其表示为用户n的β n={β 1,β 2,...,β l n},对应于用户n自身的喜好集合向量l n,β i=1,2,...表示群组中需求d ni的总数。
本实施例通过用户n的社会价值来反映用户与群组之间的社交关系,将用户n的社会价值表示为:
Figure PCTCN2021079866-appb-000021
重复需求的成本由相应的用户平均承担,其他需求则由其自己支付。否则,它就无法加入群组,只能自己承担所有成本。
用户n所在群组a n的成本可以基于下式计算得到:
Figure PCTCN2021079866-appb-000022
式中,l n表示为用户n自身的喜好集合l n=[d n1,d n2,d n3,...,d nl n],d nj∈D total,1≤j≤l n;D total=[d 1,d 2,d 3,...,d D max]表示为群组a n所有成员的喜好集合;β i表示群组a n中需求d ni的总数量;J n表示用户n的邻居集,a n表示用户n所在群组,a Jn表示用户n的邻居集所在群组;a 0表示偏好成本,l c表示与群组a n内其他成员相同的偏好,L max表示群组a n中的所有偏好。
根据社交关系模拟群组成员为了分担成本而聚集为群组的选择,表示为:M={S,v,F,A,P},其中,S表示决策用户集,v表示效用函数,F表示决定用户加入或退出群组的决策函数,A={a 1,a 2,…,a N}表示总用户决策集,a n表示群组n,P表示群组结构P={CO 1,CO 2,…,CO m};CO m表示群组m的结构特征,将效用函数v定义为用户所在其当前群组的成本,表示为:
u n(a n,a -n)=r n(a n,a Jn)     (3)
通过形成一个更好的群组结构来降低每个用户的成本,以此来达到稳定划分的效果。
步骤2:成员会选择收益更高的群组,自私偏好顺序使得群组成员自私的选择利益最大化的群组,以此来谋求更高的利益,因此通过自私偏好顺序制定游戏规则模拟群组成员的选择,具体的:
对于用户n、用户n当前所在群组a n和基于自私偏好顺序将来选择加入的群组
Figure PCTCN2021079866-appb-000023
群组a n和群组
Figure PCTCN2021079866-appb-000024
用户n并不关心群组a n中其他用户的利益,它只会选择收益更高的群组,基于自私偏好顺序选择利益最大化的群组,以此来谋求更高的利益,将该过程可表示为:
Figure PCTCN2021079866-appb-000025
式中,r n(a n)表示用户n在群组a n中的收益,
Figure PCTCN2021079866-appb-000026
表示用户n在群组
Figure PCTCN2021079866-appb-000027
中的收益,r i(a n)表示群组a n的整体收益。
群组a n和群组
Figure PCTCN2021079866-appb-000028
中的用户均具有可观的收益,记用户当前所在群组为群组a n,若当用户n基于自私偏好顺序选择加入群组
Figure PCTCN2021079866-appb-000029
时,可降低群组
Figure PCTCN2021079866-appb-000030
的整体成本且群组a n中的其他用户的成本不会增加,则可将群组a n和群组
Figure PCTCN2021079866-appb-000031
合并成一个新的群组,将以上过程定义为合并规则,并表示如下:
Figure PCTCN2021079866-appb-000032
记用户当前所在群组为群组a n,若用户脱离群组a n后,可获得较低的成本,则将群组a n进行拆分成两部分,将拆分后用户所在群组记为
Figure PCTCN2021079866-appb-000033
另一部分内的其他用户根据其自私偏好顺序选择加入其他群组,将以上过程定义为拆分规则,并表示如下:
Figure PCTCN2021079866-appb-000034
步骤3:从用户的角度回顾群组的形成过程,群组的形成过程可模拟为确切的潜在博弈,而潜在博弈至少具有一个纳什均衡点,根据潜在博弈理论得到步骤2的划分过程是有穷的,当达到纳什均衡点群组成员都没有动机离开群组达到稳定划分的作用,通过寻找纳什均衡点可以达到对群组稳定划分的效果。
现对群组的形成过程可模拟为确切的潜在博弈的说明如下:
当用户单方面更改其决策时,该潜在函数差的变化和其效用函数差的变化相同,则说明步骤2的划分过程是具有潜在函数的精确潜在博弈,即表示为:
Figure PCTCN2021079866-appb-000035
根据定义确切的潜在博弈总是具有纳什均衡,本发明的潜在函数是所有用户的社会价值的总和,表示如下:
Figure PCTCN2021079866-appb-000036
通常,较高的社会价值总会带来较高的收益,社会价值反映了群组收益,因此,使用社会价值这一指标。
假设用户n单方面将决策从群组a n离开到群组
Figure PCTCN2021079866-appb-000037
S n(a n,a -n)表示离开群组a n用户n的社会价值变化,
Figure PCTCN2021079866-appb-000038
表示加入的群组
Figure PCTCN2021079866-appb-000039
用户n的社会价值的变化:
Figure PCTCN2021079866-appb-000040
式中,a -n表示用户n离开群组a n后的群组,
Figure PCTCN2021079866-appb-000041
表示用户n尚未加入群组
Figure PCTCN2021079866-appb-000042
的群组,l n表示为用户n自身的喜好集合l n=[d n1,d n2,d n3,...,d nl n],d nj∈D total,1≤j≤l n;β i表示群组a n中需求d ni的总数量;β′ i表示群组中需求d ni的总数量。
势函数的变化为:
Figure PCTCN2021079866-appb-000043
其中,N表示用户集,该用户集分为四个部分:用户n、群组a n中的其余用户记为J an,加入的群组
Figure PCTCN2021079866-appb-000044
中的其余用户记为
Figure PCTCN2021079866-appb-000045
和剩余用户。
因为用户n离开群组a n,群组a n中的其余用户的社会价值下降,其下降恰好等于用户n的社会价值。由于用户n加入群组
Figure PCTCN2021079866-appb-000046
群组
Figure PCTCN2021079866-appb-000047
中的其余用户的社会价值有所增加,其增加恰好等于用户n的社会价值剩余用户的社会价值没有改变。以上各社会价值变化数学表达如下:
Figure PCTCN2021079866-appb-000048
Figure PCTCN2021079866-appb-000049
Figure PCTCN2021079866-appb-000050
因此,势函数的变化为:
Figure PCTCN2021079866-appb-000051
综上,群组形成过程是确切的潜在博弈,保证了本方法对群组划分的稳定 性。
根据精确潜在博弈的有限改进性质,潜在博弈中的最佳响应算法始终收敛于纳什均衡。根据纳什均衡定理可知:
NA=(NA 1,...,NA |G|)    (13)
Figure PCTCN2021079866-appb-000052
当群组划分达到纳什均衡点后,群组的成员不再有动机离开自身的群组。这样群组的划分就达到了稳定的状态,群组的收益也得到了最大化,对于群组的每个成员来说现在的结果都是可接受的。
本实施例还提供了一种融合社交关系与自私偏好顺序的群组划分系统,所述融合社交关系与自私偏好顺序的群组划分系统包括:
初步群组构建模块,用于根据用户的社交价值模拟各用户为了分担成本而聚集为群组的选择,形成初步群组。
稳定群组构建模块,用于通过自私偏好顺序制定合并和拆分规则,并基于合并和拆分规则对步骤1得到的群组进行合并或拆分,当达到纳什均衡点时,群组合并或划分结束,得到稳定群组。其中,所述的根据用户的社交价值模拟各用户为了分担成本而聚集为群组的选择,将该过程表示为:M={S,v,F,A,P},其中,S表示决策用户集,v表示效用函数,F表示决定用户加入或退出群组的决策函数,A={a 1,a 2,…,a N}表示总用户决策集,a n表示群组n,P表示群组结构P={CO 1,CO 2,…,CO m};CO m表示群组m的结构特征,所述效用函数定义为用户所在其当前群组的成本。
可选地,所述成员所在群组a n的成本表示为:
Figure PCTCN2021079866-appb-000053
式中,l n表示为用户n自身的喜好集合l n=[d n1,d n2,d n3,...,d nl n],d nj∈D total,1≤j≤l n;D total=[d 1,d 2,d 3,...,d D max]表示为群组a n所有成员的喜 好集合;β i表示群组a n中需求d ni的总数量;J n表示用户n的邻居集,a n表示用户n所在群组,a Jn表示用户n的邻居集所在群组;a 0表示偏好成本,l c表示与群组a n内其他成员相同的偏好,L max表示群组a n中的所有偏好;
所述效用函数表示为:
u n(a n,a -n)=r n(a n,a Jn)
式中,a -n表示用户n离开群组a n后的群组。
可选地,所述用户n的社交价值表示为:
Figure PCTCN2021079866-appb-000054
可选地,所述自私偏好表示为:
Figure PCTCN2021079866-appb-000055
式中,a n表示用户当前所在群组,
Figure PCTCN2021079866-appb-000056
表示用户基于自私偏好顺序将来选择加入的群组,群组a n和群组
Figure PCTCN2021079866-appb-000057
r n(a n)表示用户在群组a n中的收益,
Figure PCTCN2021079866-appb-000058
表示用户在群组
Figure PCTCN2021079866-appb-000059
中的收益,r i(a n)表示群组a n的整体收益;
所述合并规则表示为:
Figure PCTCN2021079866-appb-000060
所述拆分规则表示为:
Figure PCTCN2021079866-appb-000061
可选地,所述纳什均衡点的获得步骤包括:
群组形成过程为确切的潜在博弈,采用最佳响应算法使潜在博弈中的潜在函数收敛,得到纳什均衡点;
所述潜在博弈的潜在函数为所有用户的社会价值的总和,表示为:
Figure PCTCN2021079866-appb-000062

Claims (10)

  1. 一种融合社交关系与自私偏好顺序的群组划分方法,其特征在于:包括以下步骤:
    步骤1:根据用户的社交价值模拟各用户为了分担成本而聚集为群组的选择,形成初步群组;
    步骤2:通过自私偏好顺序制定合并和拆分规则,并基于合并和拆分规则对步骤1得到的群组进行合并或拆分,当达到纳什均衡点时,群组合并或划分结束,得到稳定群组;
    其中,所述的根据用户的社交价值模拟各用户为了分担成本而聚集为群组的选择,将该过程表示为:M={S,v,F,A,P},其中,S表示决策用户集,v表示效用函数,F表示决定用户加入或退出群组的决策函数,A={a 1,a 2,…,a N}表示总用户决策集,a n表示群组n,P表示群组结构P={CO 1,CO 2,…,CO m};CO m表示群组m的结构特征,所述效用函数定义为用户所在其当前群组的成本。
  2. 根据权利要求1所述的一种融合社交关系与自私偏好顺序的群组划分方法,其特征在于:所述成员所在群组a n的成本表示为:
    Figure PCTCN2021079866-appb-100001
    式中,l n表示为用户n自身的喜好集合l n=[d n1,d n2,d n3,…,d nln],d nj∈D total,1≤j≤l n;D total=[d 1,d 2,d 3,…,d Dmax]表示为群组a n所有成员的喜好集合;β i表示群组a n中需求d ni的总数量;J n表示用户n的邻居集,a n表示用户n所在群组,a Jn表示用户n的邻居集所在群组;a 0表示偏好成本,l c表示与群组a n内其他成员相同的偏好,L max表示群组a n中的所有偏好;
    所述效用函数表示为:
    u n(a n,a -n)=r n(a n,a Jn)  (3)式中,a -n表示用户n离开群组a n后的群组。
  3. 根据权利要求2所述的一种融合社交关系与自私偏好顺序的群组划分 方法,其特征在于:所述用户n的社交价值表示为:
    Figure PCTCN2021079866-appb-100002
  4. 根据权利要求1所述的一种融合社交关系与自私偏好顺序的群组划分方法,其特征在于:所述自私偏好表示为:
    Figure PCTCN2021079866-appb-100003
    式中,a n表示用户当前所在群组,
    Figure PCTCN2021079866-appb-100004
    表示用户基于自私偏好顺序将来选择加入的群组,群组a n和群组
    Figure PCTCN2021079866-appb-100005
    r n(a n)表示用户在群组a n中的收益,
    Figure PCTCN2021079866-appb-100006
    表示用户在群组
    Figure PCTCN2021079866-appb-100007
    中的收益,r i(a n)表示群组a n的整体收益;
    所述合并规则表示为:
    Figure PCTCN2021079866-appb-100008
    所述拆分规则表示为:
    Figure PCTCN2021079866-appb-100009
  5. 根据权利要求3所述的一种融合社交关系与自私偏好顺序的群组划分方法,其特征在于:所述步骤2中纳什均衡点的获得步骤包括:
    群组形成过程为确切的潜在博弈,采用最佳响应算法使潜在博弈中的潜在函数收敛,得到纳什均衡点;
    所述潜在博弈的潜在函数为所有用户的社会价值的总和,表示为:
    Figure PCTCN2021079866-appb-100010
  6. 一种融合社交关系与自私偏好顺序的群组划分系统,其特征在于,所述融合社交关系与自私偏好顺序的群组划分系统包括:
    初步群组构建模块,用于根据用户的社交价值模拟各用户为了分担成本而聚集为群组的选择,形成初步群组;
    稳定群组构建模块,用于通过自私偏好顺序制定合并和拆分规则,并基于合并和拆分规则对步骤1得到的群组进行合并或拆分,当达到纳什均衡点时,群组合并或划分结束,得到稳定群组;
    其中,所述的根据用户的社交价值模拟各用户为了分担成本而聚集为群组的选择,将该过程表示为:M={S,v,F,A,P},其中,S表示决策用户集,v表示效用函数,F表示决定用户加入或退出群组的决策函数,A={a 1,a 2,…,a N}表示总用户决策集,a n表示群组n,P表示群组结构P={CO 1,CO 2,…,CO m};CO m表示群组m的结构特征,所述效用函数定义为用户所在其当前群组的成本。
  7. 根据权利要求6所述的融合社交关系与自私偏好顺序的群组划分系统,其特征在于,所述成员所在群组a n的成本表示为:
    Figure PCTCN2021079866-appb-100011
    式中,l n表示为用户n自身的喜好集合l n=[d n1,d n2,d n3,…,d nln],d nj∈D total,1≤j≤l n;D total=[d 1,d 2,d 3,…,d Dmax]表示为群组a n所有成员的喜好集合;β i表示群组a n中需求d ni的总数量;J n表示用户n的邻居集,a n表示用户n所在群组,a Jn表示用户n的邻居集所在群组;a 0表示偏好成本,l c表示与群组a n内其他成员相同的偏好,L max表示群组a n中的所有偏好;
    所述效用函数表示为:
    u n(a n,a -n)=r n(a n,a Jn)
    式中,a -n表示用户n离开群组a n后的群组。
  8. 根据权利要求7所述的融合社交关系与自私偏好顺序的群组划分系统,其特征在于,所述用户n的社交价值表示为:
    Figure PCTCN2021079866-appb-100012
  9. 根据权利要求6所述的融合社交关系与自私偏好顺序的群组划分系统,其特征在于,所述自私偏好表示为:
    Figure PCTCN2021079866-appb-100013
    式中,a n表示用户当前所在群组,
    Figure PCTCN2021079866-appb-100014
    表示用户基于自私偏好顺序将来选择加入的群组,群组a n和群组
    Figure PCTCN2021079866-appb-100015
    r n(a n)表示用户在群组a n中的收益,
    Figure PCTCN2021079866-appb-100016
    表示用户在群组
    Figure PCTCN2021079866-appb-100017
    中的收益,r i(a n)表示群组a n的整体收益;
    所述合并规则表示为:
    Figure PCTCN2021079866-appb-100018
    所述拆分规则表示为:
    Figure PCTCN2021079866-appb-100019
  10. 根据权利要求8所述的融合社交关系与自私偏好顺序的群组划分系统,其特征在于,所述纳什均衡点的获得步骤包括:
    群组形成过程为确切的潜在博弈,采用最佳响应算法使潜在博弈中的潜在函数收敛,得到纳什均衡点;
    所述潜在博弈的潜在函数为所有用户的社会价值的总和,表示为:
    Figure PCTCN2021079866-appb-100020
PCT/CN2021/079866 2020-04-03 2021-03-10 一种融合社交关系与自私偏好顺序的群组划分方法 WO2021197001A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US17/795,453 US20230091117A1 (en) 2020-04-03 2021-03-10 Group division method based on social relationship in combination with selfish preference order

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202010259534.1A CN111488532B (zh) 2020-04-03 2020-04-03 一种融合社交关系与自私偏好顺序的群组划分方法
CN202010259534.1 2020-04-03

Publications (1)

Publication Number Publication Date
WO2021197001A1 true WO2021197001A1 (zh) 2021-10-07

Family

ID=71810838

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/079866 WO2021197001A1 (zh) 2020-04-03 2021-03-10 一种融合社交关系与自私偏好顺序的群组划分方法

Country Status (3)

Country Link
US (1) US20230091117A1 (zh)
CN (1) CN111488532B (zh)
WO (1) WO2021197001A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113992614A (zh) * 2021-10-26 2022-01-28 广州博冠信息科技有限公司 会话群组的处理方法及装置、计算机存储介质、电子设备

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111488532B (zh) * 2020-04-03 2022-10-14 南京邮电大学 一种融合社交关系与自私偏好顺序的群组划分方法
CN112822430B (zh) * 2020-12-29 2022-12-30 上海掌门科技有限公司 一种会议群组合并的方法与设备
CN117454022A (zh) * 2023-10-31 2024-01-26 南昌大学 一种基于人工智能物联网的隐式组推荐方法及系统

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20180300821A1 (en) * 2012-03-06 2018-10-18 Immersonal, Inc. Group event or activity recommendations via social-relationship-related override conditions
CN110209954A (zh) * 2019-06-03 2019-09-06 南京邮电大学 基于lda主题模型和深度学习的群组推荐方法
CN111488532A (zh) * 2020-04-03 2020-08-04 南京邮电大学 一种融合社交关系与自私偏好顺序的群组划分方法

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20110087679A1 (en) * 2009-10-13 2011-04-14 Albert Rosato System and method for cohort based content filtering and display
CN107391637B (zh) * 2017-07-10 2021-01-26 江苏省现代企业信息化应用支撑软件工程技术研发中心 针对具备地理社交信息的群组推荐方法
CN109871482A (zh) * 2019-01-15 2019-06-11 南京邮电大学 一种基于纳什均衡的群组教育资源推荐方法

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20180300821A1 (en) * 2012-03-06 2018-10-18 Immersonal, Inc. Group event or activity recommendations via social-relationship-related override conditions
CN110209954A (zh) * 2019-06-03 2019-09-06 南京邮电大学 基于lda主题模型和深度学习的群组推荐方法
CN111488532A (zh) * 2020-04-03 2020-08-04 南京邮电大学 一种融合社交关系与自私偏好顺序的群组划分方法

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
GUO ZHIWEI: "Research on Group-Based Recommendation via Multi-Feature Fusion in Social Networks", ELECTRONIC TECHNOLOGY & INFORMATION SCIENCE, CHINA DOCTORAL DISSERTATIONS FULL-TEXT DATABASE, 15 April 2019 (2019-04-15), XP055855526 *
HOU FUJUN; ZHAI YUBING; YOU XINLI: "An equilibrium in group decision and its association with the Nash equilibrium in game theory", COMPUTERS & INDUSTRIAL ENGINEERING, PERGAMON, AMSTERDAM, NL, vol. 139, 31 October 2019 (2019-10-31), AMSTERDAM, NL, XP086042359, ISSN: 0360-8352, DOI: 10.1016/j.cie.2019.106138 *
WU YUNCHANG, BAISONG LIU, YANGYANG WANG, CHENJIE FEI: "Review of group recommendation analysis and research", TELECOMMUNICATIONS SCIENCE, 31 December 2018 (2018-12-31), pages 71 - 83, XP055855522, DOI: 10.11959/j.issn.1000−0801.2018306 *
ZENG XIANLIN, JIE CHEN, SHU LIANG, YIGUANG HONG: "Generalized Nash equilibrium seeking strategy for distributed nonsmooth multi-cluster game", AUTOMATICA, vol. 103, 1 May 2019 (2019-05-01), pages 20 - 26, XP055855536, DOI: 10.1016/j.automatica.2019.01.025 *

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113992614A (zh) * 2021-10-26 2022-01-28 广州博冠信息科技有限公司 会话群组的处理方法及装置、计算机存储介质、电子设备

Also Published As

Publication number Publication date
CN111488532A (zh) 2020-08-04
CN111488532B (zh) 2022-10-14
US20230091117A1 (en) 2023-03-23

Similar Documents

Publication Publication Date Title
WO2021197001A1 (zh) 一种融合社交关系与自私偏好顺序的群组划分方法
Yu et al. A multi-criteria decision-making model for hotel selection with linguistic distribution assessments
Ganley et al. The ties that bind: Social network principles in online communities
Yu et al. Hierarchical punishment-driven consensus model for probabilistic linguistic large-group decision making with application to global supplier selection
CN111125453B (zh) 基于子图同构的社交网络中意见领袖角色识别方法及存储介质
Han et al. Pricing and bargaining strategy of e-retail under hybrid operational patterns
CN113761383B (zh) 一种基于统一用户行为建模的搜索和推荐融合系统
Xie et al. Effects of host incentives on multiple listings in accommodation sharing
CN107145541B (zh) 基于超图结构的社交网络推荐模型构建方法
CN111324807A (zh) 基于信任度的协同过滤推荐方法
CN109840702A (zh) 一种基于多核融合的新项目协同推荐方法
Chen et al. Dim-ds: Dynamic incentive model for data sharing in federated learning based on smart contracts and evolutionary game theory
Tang et al. Combining global and local trust for service recommendation
Ruiz Measuring the joint distribution of household's income, consumption and wealth using nested Atkinson measures
Song et al. Day ahead bidding of a load aggregator considering residential consumers demand response uncertainty modeling
Huang et al. Pareto approach for DEA cross efficiency evaluation based on interval programming
CN110570300A (zh) 基于XGBoost驱动的海萨尼转换金融决策处理系统
Sun et al. Dynamic optimal pricing strategies for knowledge-sharing platforms
CN113051487A (zh) 基于用户信任和社交影响力的旅游群组推荐方法
CN114841497A (zh) 考虑非合作行为的大群体卫星应急方案决策方法和系统
Coucheney et al. Influence of search neutrality on the economics of advertisement-financed content
Lin et al. GRHAM: towards group recommendation using hierarchical attention mechanism
Yan Prior-independence: A new lens for mechanism design
Chen et al. Optimize emission reduction commitments for international environmental agreements
Li et al. Credit expansion and concentration in China based on GMCR

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: 21780941

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: 21780941

Country of ref document: EP

Kind code of ref document: A1