A Hybrid Preference-based Recommender System Based on Fuzzy Concordance/Discordance Principle

摘要

A variety of filtering techniques have been proposed to deal with the huge amount of data gathered by today's Web sites thereby reducing the impact of information overload problem. Despite their strengths, each filtering technique has its own weaknesses leading to hybrid filtering techniques to overcome individual weaknesses. Moreover, the aggregation used by most filtering techniques is compensatory and allows very high (pros) and very low (cons) scores to compensate each other. This is not always suitable, especially when criteria are very conflicting and/or the information about alternatives is incomplete. Another approach is to measure the pros and cons separately and then the overall statement about a given pair of alternatives is obtained by balancing the pros and cons within the set of criteria. In this paper we propose a hybrid collaborative/demographic filtering technique that implements a preference-based filtering using a fuzzy concordance/discordance principle. This hybridization helps new users and all those users who have rated less number of items to get assistance from users having close demographical attributes. The experimental results reveal that our approach outperforms the classical approach.