Study on User Behavior Prediction Based on Singular Value Decomposition
Yongxia Jing, Heping Gou, Fu Chuanyi, Qiang Liu · 2016
Social network, such as micro-blog, includes the large amount of users' data which reflects the user behavior. The users' behavior, such as emotion and interests can be analyzed effectively through data mining and other means. However, due to the posts, comment and others data in social network are relatively short, the huge dimensionality of the item-document matrix which is built according to these users' data is getting larger and larger with increasing of the number of users and data. In order to improve the efficiency of data analysis, one of the important methods is to reduce the dimensionality of the data matrix. In the paper, singular value decomposition method is used to reduce the data dimensionality, the data matrix with large dimensionality can be mapped to another semantic space with low dimensionality, reducing the cost of the similarity computing between documents. The documents, such as blogs, posts and comments which different users posted, can be divided to different topics, the number of documents in the topics can reflect the state of user emotion and behavior. Experiments show effectiveness of the proposed method.