A Robust Multi-Criteria Collaborative Filtering Algorithm

Abstract

Collaborative filtering recommender systems assist individuals to discover relevant products or services that they might be interested in a large set of alternatives by analyzing the collected preferences. Recent research presents that the accuracy of recommendations might be improved significantly by collecting multi-criteria user preferences. Such rating scheme allows users to express their preferences better. Multi-criteria collaborative filtering algorithms are suitable for utilizing in many domains such as research paper, movie, or hotel recommendation. However, such systems are vulnerable to shilling attacks. In order to prevent manipulations, robust recommendation algorithms are required. Although multi-criteria collaborative filtering algorithms were evaluated in several dimension, robustness against shilling attacks has not been studied as a feature. In this paper, we propose an attack-resistant multi-criteria collaborative filtering algorithm. Experimental evaluation confirms that the proposed algorithm is not deeply affected against all known attack types.