Personalized recommendation based on WAR

Abstract

Personalized recommendation is the process of customizing Web site to meet the requirement of specific user by analyzing user's browsing behavior and extracting knowledge based on Web logs. Association rules mining technology is widely adopted in this field. However, the resulting association patterns can not effectively predict user's future browsing behavior because of the low similarity degree between resulting rules and user's browsing behavior. In this paper, we assign a weight to each item in a transaction to reflect the interest degree, which extends the traditional association rule method. We also proposed weighted association rule (WAR) through associating a weight with each item in resulting association rules. Each Web page is assigned to a weight according to interest degree and three key factors, i.e. visit frequency, stay duration and operation time. A novel personalized recommendation mechanism is presented based our proposed WAR. The weighted measurement in our personalized recommendation can be used to determine the importance of Web pages for user. We try to acquire user's requirement more precisely so as to more useful Web pages are discovered and recommended for user.