基于贝叶斯网络的用户行为相似性研究

Research on Users Behavior Similarity Based on Bayesian Network

摘要: 随着移动设备和移动服务的高速发展，移动社交网络融入了人们的日常生活。每时每刻人们都在这里生成大量的数据，而对于这个巨大的社交媒体数据源的研究是非常有意义和必要的。但在对社交网络的数据挖掘中，发现存在大量的不确定性，以社交网络中的最广为人知的推荐算法为例，如何利用已知的用户信息为该用户更为精准地推荐其感兴趣的信息，这其中就蕴藏着大量的不确定性，如何清楚地展示和度量用户相似性这种不确定性知识，在商品推荐和社交网络用户关系演化等研究中一直是艰巨的挑战。因此本文提出采用贝叶斯网络这一结合拓扑结构和概率推理的重要的概率图模型作为发现这种不确定知识的框架，并基于此提出了一种用户相似性发现方法。

Abstract: With the rapid development of mobile devices and mobile services, mobile social networks are integrated into people’s daily lives, and people are also generating a large amount of data here. The research on this huge data source is very meaningful and necessary. User similarity in social networks is an important research field in social media data analysis. It also plays a very important role in the research of product recommendation and social network user relationship evolution. The similarity between users depends not only on the network topology, but also on the degree of dependence between users. In order to achieve the similarity measure between users in social network data, this paper proposes a basis based on topology and probabilistic reasoning. The user similarity measurement method of social network is adopted, and Bayesian network is used as the framework of this uncertain knowledge discovery. A user similarity discovery method based on Bayesian network is proposed.