Improving the Performance of User-Based Collaborative Filtering by Mining Latent Attributes of Neighborhood

Abstract

In the area of recommender systems, user-based collaborative filtering algorithm has been extensively studied and discussed. In the traditional approach of this method, a target user's preference for an item is predicted by the integrated preference of the user's neighbors for the item, ignoring the structure of these neighbors. That is, these neighbors form two distinct groups: some neighbors may like the target item or give high rating, on the other hand, some neighbors may dislike the target item or give low rating. The structure of the two groups may influence user's choice. As an extension of user-based collaborative filtering, this paper focuses on the analysis of such structure by mining latent attributes of users' neighborhood, and corresponding correlations with users' preference by several popular data mining techniques. Mining latent attributes and experiment evaluation was conducted on Movie Lens data set. The experimental results reveal that the proposed method can improve the performance of pure user-based collaborative filtering algorithm.