Personalized Recommendation Based On Behavior Model

With the development of technology, the Internet has become the largest information database in today’s world. Faced with such a large amount of information, how to make use of these data is the focus of current research. Personalized information recommendation is a new re-search field emerged with the increasingly use of data mining techniques which analyzes the preferences of users according to the access behavior to provide personalized recommendation service when users access the web site, and to improve customer satisfaction and loyalty. There are many kinds of algorithms for making predictions for the target users, and among them Collaborative Filter-ing (CF) is widely adopted. In some domains, a user’s behavior sequences reflect his/her preferences over items so that users who have similar behavior sequences may indicate they have similar prefer-ence models. Based on this fact, we discuss how to improve the col-laborative filtering algorithm by using user behavior sequence similarity. We proposed an new Behavior Sequence Similarity Measurement (BSSM) approach. Then, different ways to combine BSSM with CF algorithm are presented. Experiments on two real test data sets prove that more precise and stable recommendation perfor-mances can be achieved.