An efficient framework to combat multidimensionality in multi-criteria recommender systems

An efficient framework to combat multidimensionality in multi-criteria recommender systems

연구 분야: Verification

학회: Cluster Computing

Recommender Systems (RS) are widely adopted software tools for easing information overload and generating personalized recommendations to the users. Collaborative filtering (CF) is one of the extensively implemented recommendation techniques that provides recommendations based on the preferences of other like-minded users. Traditional RS generally works on a single numerical rating on items provided by the users. Recent research suggests that the incorporation of multi-criteria ratings into classical RS has greatly improved the utility of recommendations. However, incorporation of multi-criteria ratings into classical systems is still a matter of concern as it causes the multidimensionality issue. In this paper, we propose an efficient framework to reduce the multidimensionality in multi-criteria recommender systems (MCRS) using two different multi-criteria aggregation methods. Unlike the traditional aggregation methods, our utilized aggregation methods combine the multiple ratings by preserving the importance of individual criteria ratings. Moreover, in order to identify the most accurate neighborhood set for each user, we combine the separately calculated overall and aggregated score based similarities together to obtain the total similarity. Extensive experiments performed on benchmark dataset indicate that our proposed approaches outperformed the existing multi-criteria recommendation approaches on various evaluation measures.