A novel multi-criteria user profile based on criteria - ranking for movie recommender

Abstract

This thesis aims at proposing a novel methodology for Movie Recommender System based on multi-criteria ratings. Many researchers have been developed the multiple criteria ratings on various ways. Most of them directly use the values of user profiles to find the relation among users, which may cause problems and tend to provide poor quality neighbors. Therefore, I propose Criteria-Ranking with Closeness Score to indicate the significance of criteria toward each user characteristic that will improve the quality of user profiles. According to the experimental evaluation, Criteria-Ranking with Closeness Score provides more accurate recommendation results than other traditional multi-criteria recommender techniques.