Abstract
As the World Wide Web becomes an important information source in recent years, it is hard to get useful knowledge from the sea of information. In this paper, we describe a technique for making personalized recommendation on the Internet environment. We introduce the concepts of the user interest and behavior that are generated from the categories of documents read by the users. A new method for mining the user interest and behavior from the documents read by the user is proposed. According to the patterns of user interests and behaviors, we define six types of web users and describe a similarity measure of the patterns to classify the users into clusters. In addition, four kinds of recommendations based on the clustering results are provided for the users. Finally, we make several experiments to evaluate the effectiveness and efficiency of our approach.