A Particle Swarm Optimization Approach to Multi Criteria Recommender System Utilizing Effective Similarity Measures

Abstract

Recommender system (RS), a web personalization tool, attempts to generate suitable recommendations to users based on their preferences. Generally, recommender system works on overall ratings but these ratings do not reflect the actual user preferences. Therefore, incorporation of multiple criteria ratings into RS can capture the user preferences accurately and produce effective recommendations to users. Multi criteria recommender systems (MCRS) generate recommendations to users based on the aggregation of similarities computed on multiple criteria using collaborative filtering. However, capturing optimal weights of various users on different criteria in the process of similarity aggregation is a major concern. Further selection of appropriate similarity measure is another challenge for employing collaborative filtering. Our work in this paper is an attempt towards developing multi criteria recommender systems by utilizing various similarity measures and particle swarm optimization to learn optimal weights. Experimental results reveal that our proposed approaches outperform other traditional approaches.