An Improved Recommendation Approach based on User's Standard and Interests

Abstract

the field of web personalization is growing exponentially. From e-mail, e-trading, and internet forums to social networking based websites, directly or indirectly utilize the concept of web personalization and recommendation system for providing customized services and attention- grabbing offers to their users. By now, a wide range of recommendation systems have been proposed by various researchers and still the research is going on so as to attain the user's expectations. It is a general thought that a user should be recommended the best product from his/her favorite product categories. In this work, we show theoretically that a recommendation system that follows the above criterion does not recommend appropriate products to all types of users. In addition, we propose a user's standard based recommendation system that overcomes this limitation of conventional approach, and provides improved recommendations covering all types of users. For evaluation purpose, we have employed the KDD MovieLens dataset and developed a movie recommendation system based on the proposed approach. We introduce the term 'Gain' for measuring the difference created by the proposed approach as compared to conventional approach. Experimental results reveal that the proposed approach benefited 56% of users and improved 23% of total recommendations. General Terms Web Usage Mining and recommendation system Keywordsmining, KDD dataset, recommendation system, web personalization