New hybrid web personalization framework

Abstract

Web personalized recommender systems based on web mining try to mine users' behavior patterns from web access logs and site metadata, and recommend pages to the online user by matching the user's browsing behavior with the mined previous user's behavior patterns. Recommendation approaches proposed in previous works, however, cannot still satisfy users especially in huge and dynamic web sites. To provide recommendation efficiently, we advance a framework for web mining-based personalization that combines web usage data with web content and site structure for predicting users' future requests more accurately. The experimental results on real dataset show that the approach can improve accuracy and coverage of recommendations to users.