Behavior Based User Interests Extraction Algorithm

Abstract

The most important issue in personalized services is how to build a proper user model for individual user. Extracting user interests is an essential question, which attracts many researchers' investigation. But most of them don't consider enough the history of user behaviors. This paper proposes a method to extract user interests from user behavior history and document information, so called Behavior based User Interests extraction (BUIE) algorithm. By analyzing document information of English scientific literatures with domain ontology, the interest items of the documents can be extracted. Then combining with the user behavior history, the user interest items' weights can be refined. Experimental results show that, BUIE algorithm could get better performance on user interests extraction than those that don't consider user behaviors. The user interests calculated by this method are quite consistent with user's real interests. The different kinds of behaviors' contribution degrees to user interests are analyzed, and it's found that the behaviors of copying, saving, printing are important factors in extracting user interests. The contribution degree of printing is higher than that of saving and copying. But the stability of these three behaviors is less than that of relative browsing time.