Multi-Criteria Ranking for Recommender Systems

Recommender Systems have been served in multiple areas (e.g., e-commerce, hotel bookings, job seeking, social media, online streaming, etc.) to assist users’ decision making. Multi-Criteria Recommender Systems (MCRS) are a special type of recommender systems, where the system can produce item recommendations by taking advantage of user preferences on different aspects of the items (i.e., criteria). Apparently, there is a process of multi-criteria decision making involved. However, the development of MCRS algorithms failed to fill the gap between multi-criteria decision making and recommender systems. Most of the MCRS algorithms simply utilized an aggregation function (e.g., weighted sum) to estimate the overall preference by a user on an item. In this talk, we discuss our recent and promising technique, multi-criteria ranking, which reuses the knowledge and techniques in multi-criteria decision making and multi-objective optimization to boost the performance of multi-criteria recommender systems, in order to deliver more accurate item recommendations.