基于非负矩阵分解的用户话题兴趣度算法

User-Topic Interestingness Algorithms Based on Non-negative Matrix Factorization

Web媒体被公认为继报纸、广播、电视之后的"第四媒体"。而Web2.0的迅速普及,又使当今的Web媒体呈现了一种"自媒体"形式,即每个用户既是信息的接受者,也是信息发布者和信息转发者,在信息传递过程中,用户与用户互动,影响信息传播的进程。用户本身的特性对于传播有很大影响,信息传播依赖于用户个体的行为模式。因此,需要对用户和传播话题之间的关系进行建模,来度量用户对某个话题的感兴趣程度。论文提出了有效的算法来对用户进行感兴趣的话题推荐,该算法基于非负矩阵分解理论,分析用户发表过的内容,将用户感兴趣的话题推荐给该用户。该文针对研究小组下载的真实数据集-科学网数据集进行实验分析,实验结果表明算法能够有效地将用户感兴趣的话题推荐给用户。

Web media is generally acknowledged as ＂the fourth media＂ after the newspaper, broadcast and TV. And as Web 2. 0 prevails over the intemet, the web media has a form called ＂self-media＇, which means that every individual is a receiver, also a publisher and forwarder at the same time. The characters of users have great influence on the process of irdormation diffusing, so a model should be built to measure the interestingness between the users and the topics. In this paper, some effective algorithms have been proposed to recommend topics to users, such algorithms are based on non-negative matrix factorization theory, these algorithms analyze the contents that users have published, and recommend topics to users. Experiments based on real word datasets, sciencenet. eom datasets show these algorithms can recommend users＇ interested topics to users.