社会网络中社区发现与用户推荐算法研究

发布时间：2018-03-27 05:24

本文选题：关联规则　切入点：相似度　出处：《新疆大学》2017年硕士论文

【摘要】：随着现实人类社会与虚拟网络融合的不断深化,新兴的社会网络为信息的快速传播和交流共享提供了便捷有效的平台。这使得社会网络中的用户能随时了解,并参与信息的交互过程。人们在信息交互环境中所留下的数字足迹形成了各种各样的关系网络。由于网络中用户关系数据具有潜在的商业及研究价值,社会网络中社区划分及用户推荐的相关研究已成为目前研究人员广泛关注的课题。本文主要研究了关联规则算法、基于线图的重叠社区发现在微博中的应用以及用户推荐算法。具体工作内容和研究成果如下:1.对于传统Apriori算法存在时间复杂度较高,无关冗余项较多的问题,提出基于位图数据和散列函数的BHA算法,结合相关性质,有效提升了算法的计算效率。同时,在基于微博用户的推荐应用中,由关注关系计算用户间的出,入相似度,并引入信任度概念,弥补了相似度未考虑用户间距离的问题。在此基础上,结合BHA算法进行用户推荐。基于微博用户数据的实验结果表明,关联规则挖掘效率及算法推荐的有效性均有所提高。2.提出了基于用户关注相似和标签相似的微博社区发现网络模型,将CNM算法与该模型相结合进行微博社区发现。对于检测结果中存在的节点过重叠现象,引入核心链路概念,对网络中的孤立边进行删除,有效解决了节点的过度重叠问题。在此基础上,进一步分析共邻好友间的关系,提出了基于社区检测的微博用户推荐算法。将所提算法应用于真实网络和微博网络,实验表明社区发现的精确度得到进一步提高,同时,将社区检测与用户推荐相结合有效提升了推荐结果的准确性。
[Abstract]:With the deepening of the integration of real human society and virtual network, the emerging social network provides a convenient and effective platform for the rapid dissemination and exchange and sharing of information, which enables the users in the social network to understand it at any time. And participate in the process of information interaction. The digital footprint left by people in the information interactive environment forms a variety of relational networks. Because of the potential commercial and research value of user relationship data in the network, The research on community division and user recommendation in social network has become a topic that researchers pay more attention to. This paper mainly studies the algorithm of association rules. The application of line graph based overlapping community discovery and user recommendation algorithm in Weibo. The specific work and research results are as follows: 1.The traditional Apriori algorithm has the problems of higher time complexity and more irrelevant redundant items. This paper proposes a BHA algorithm based on bitmap data and hash function, which combines the related properties and improves the efficiency of the algorithm. At the same time, in the recommended application of Weibo, the similarity between users is calculated by the concern relation. The concept of trust is introduced to make up for the problem that the similarity does not take into account the distance between users. On this basis, the BHA algorithm is used to recommend users. The experimental results based on Weibo user data show that, The efficiency of association rules mining and the effectiveness of algorithm recommendation are improved. 2. Weibo community discovery network model based on user concern similarity and label similarity is proposed. The CNM algorithm is combined with the model for Weibo community discovery. The concept of core link is introduced to remove the isolated edges in the network. The problem of excessive overlap of nodes is effectively solved. On this basis, the relationship between common neighbors and friends is further analyzed, and a recommendation algorithm for Weibo users based on community detection is proposed. The proposed algorithm is applied to real networks and Weibo networks. The experimental results show that the accuracy of community discovery is further improved, and the accuracy of recommended results is improved effectively by combining community detection with user recommendation.
【学位授予单位】：新疆大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：O157.5;TP391.3

【相似文献】