Optimizing collaborative filtering recommender systems

Abstract

Collaborative filtering (CF) is the most successful recommendation technique, which has been used in a number of different applications. In traditional CF, the ratings of all items are equally weighted when similarity measure is calculated. But, if the importance of features (or items) is different respectively, feature weighting structure needs to be changed according to the importance of features. This paper presents a CA based feature weighting method. Through this weighting method, we can focus on the good items while removing bad ones or reducing their impacts.