Feature-weighted user model for recommender systems

Recommender systems are gaining widespread acceptance in e-commerce applications to confront the "information overload" problem. Collaborative Filtering (CF) is a successful recommendation technique, which is based on past ratings of users with similar preferences. In contrast, Content-Based filtering (CB) assumes that each user operates independently. As a result, it exploits only information derived from document or item features. Both approaches have been extensively combined to improve the recommendation procedure. Most of these systems are hybrid: they run CF on the results of CB and vice versa. CF exploits information from the users and their ratings. CB exploits information from items and their features. In this paper, we construct a featureweighted user profile to disclose the duality between users and features. Exploiting the correlation between users and features we reveal the real reasons of their rating behavior. We perform experimental comparison of the proposed method against the well-known CF, CB and a hybrid algorithm with a real data set. Our results show significant improvements, in terms of effectiveness.