社交网络中基于用户偏好变化的影响最大化研究
发布时间：2018-11-12 18:53
【摘要】：随着信息技术和互联网技术的发展,诸如Facebook、微信等具有社交功能网站获得了巨大的成功。影响最大化问题旨在挖掘社交网络中最有影响力的Top-k个节点的集合,是社交网络研究领域中的关键问题。但是一个用户在一个社交网络中可能对不同的话题感兴趣,且偏好度不同,同时随着时间的推移,用户对话题的偏好度也会发生变化。在之前的很多工作中,影响最大化问题都忽视了这些因素,所挖掘的用户都是全局模式下最有影响力的用户。如果我们需要查找当前特定话题下最有影响力的用户,传统算法的精确度会受到影响。 在此背景下,本文提出了基于用户偏好变化的影响最大化问题,建立了一个考虑用户偏好变化的UCP_IC(Independent Cascade Model based on User Current Preferences)影响传播模型。在UCP_IC模型中,为了解决用户偏好变化问题,模型根据生物学中艾宾浩斯遗忘规律,设计了随时间间隔递减的指数函数来衡量用户当前对话题的偏好。此外为了将用户之间激活概率与用户偏好关联起来,我们同时考虑用户之间的在特定话题下的联系频率与用户对话题的偏好,并使用关联规则的方法将两者联系起来作为用户间激活概率。在此模型的基础上,我们提出了GAUCP(Greedy Algorithm based on User Current Preferences)算法来挖掘当前在特定话题下最有影响力的用户。该算法在考虑用户当前偏好的情况下采用了贪心算法来挖掘用户。在特定话题下,其能取得更好的精确度。基于影响传播模型的子模特性,算法结果可以获得约63%的精确度保证,并能使用CELF对算法计算效率进行优化。 最后基于DBLP学术数据库中相关数据进行了实验,在特定话题下,GAUCP可以找到当前对话题最有影响力的用户集。
[Abstract]:With the development of information technology and Internet technology, social networking sites such as Facebook, WeChat have achieved great success. Impact maximization is a key issue in the research field of social networks, which aims to excavate the set of the most influential Top-k nodes in social networks. However, a user may be interested in different topics in a social network, and their preferences may vary with the passage of time. In a lot of previous work, the problem of influence maximization ignores these factors, and the users mined are the most influential users in the global mode. If we need to find the most influential users on a given topic, the accuracy of traditional algorithms will be affected. Under this background, this paper proposes the problem of maximizing the influence of user preference change, and establishes a UCP_IC (Independent Cascade Model based on User Current Preferences) influence propagation model considering the change of user preference. In the UCP_IC model, in order to solve the problem of user preference change, according to the rule of Obinhos forgetting in biology, the exponential function of decreasing with time interval is designed to measure the user's current preference to the topic. In addition, in order to correlate the activation probability between users and user preferences, we also consider the frequency of contact between users under a specific topic and the user preference for the topic. The association rules are used to associate the two as the activation probability between users. Based on this model, we propose a GAUCP (Greedy Algorithm based on User Current Preferences) algorithm to mine the most influential users on specific topics. The greedy algorithm is used to mine users considering the current preferences of users. In a given topic, it can achieve better accuracy. Based on the influence of the submodel of the propagation model, the accuracy of the algorithm can be guaranteed by about 63%, and the computational efficiency of the algorithm can be optimized by using CELF. Finally, based on the relevant data in the DBLP academic database, GAUCP can find the most influential user set on the topic under the specific topic.
【学位授予单位】：云南大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：O157.5
本文编号：2327920
[Abstract]:With the development of information technology and Internet technology, social networking sites such as Facebook, WeChat have achieved great success. Impact maximization is a key issue in the research field of social networks, which aims to excavate the set of the most influential Top-k nodes in social networks. However, a user may be interested in different topics in a social network, and their preferences may vary with the passage of time. In a lot of previous work, the problem of influence maximization ignores these factors, and the users mined are the most influential users in the global mode. If we need to find the most influential users on a given topic, the accuracy of traditional algorithms will be affected. Under this background, this paper proposes the problem of maximizing the influence of user preference change, and establishes a UCP_IC (Independent Cascade Model based on User Current Preferences) influence propagation model considering the change of user preference. In the UCP_IC model, in order to solve the problem of user preference change, according to the rule of Obinhos forgetting in biology, the exponential function of decreasing with time interval is designed to measure the user's current preference to the topic. In addition, in order to correlate the activation probability between users and user preferences, we also consider the frequency of contact between users under a specific topic and the user preference for the topic. The association rules are used to associate the two as the activation probability between users. Based on this model, we propose a GAUCP (Greedy Algorithm based on User Current Preferences) algorithm to mine the most influential users on specific topics. The greedy algorithm is used to mine users considering the current preferences of users. In a given topic, it can achieve better accuracy. Based on the influence of the submodel of the propagation model, the accuracy of the algorithm can be guaranteed by about 63%, and the computational efficiency of the algorithm can be optimized by using CELF. Finally, based on the relevant data in the DBLP academic database, GAUCP can find the most influential user set on the topic under the specific topic.
【学位授予单位】：云南大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：O157.5
【参考文献】
相关期刊论文 前3条
1 田家堂;王轶彤;冯小军;;一种新型的社会网络影响最大化算法[J];计算机学报;2011年10期
2 陈浩;王轶彤;;基于阈值的社交网络影响力最大化算法[J];计算机研究与发展;2012年10期
3 曹玖新;董丹;徐顺;郑啸;刘波;罗军舟;;一种基于k-核的社会网络影响最大化算法[J];计算机学报;2015年02期
，本文编号：2327920
本文链接：https://www.wllwen.com/kejilunwen/yysx/2327920.html
