基于非负多矩阵分解的微博网络信息推荐*

微博网站作为一种流行的社交媒体形式，在为用户提供丰富信息和服务的同时，也带来了信息超载问题。如何利用微博网络为用户推荐有价值的信息，以缓解信息超载问题，变得日益重要。根据微博网络的有向性以及建立关注关系的随意性等特点，提出了一种基于非负多矩阵分解的微博网络推荐方法，综合考虑了用户之间的关注关系、用户与微博内容的转发关系，以及微博内容与主题的所属关系等多源信息。基于新浪微博数据集进行了微博内容推荐实验，结果表明基于非负多矩阵分解的方法，能够有效利用微博网络中的多维信息，显著提高推荐准确度。本方法不仅能挖掘出微博内容的主题，还能挖掘出用户间的关联关系，还可推广到对用户进行好友和主题的推荐。

Information recommendation in Microblogging network based on non-negative multiple matrix factorization

As a popular form of social media, micro-blog web site providing rich information and services to users, at the same time it also brings the problem of information overload. How to use the microblogging network to recommend valuable information for users, to ease the problem of information overload, become increasingly important. According to the orientation, the randomness followed and other characteristics in the microblogging network, this paper proposed a microblogging network recommendation based on non-negative multiple matrix factorization, comprehensive consideration of the follow relationship among users, the repost relationship between users and micro-blog content, and the ownership relationship between micro-blog content and topic and other multi-source information. Sina weibo dataset for the micro-blog content recommendation experiment, the results show that the method based on non-negative multiple matrix factorization, can effectively use the multidimensional information in microblogging network, significantly improve the accuracy of recommendation. This method can not only dig out the theme of micro-blog content, but also dig out the relationship among users, and can be extended to recommend friends and topics to users.