Recommentdation methodology using dynamic and hybrid user profile, and multiple criteria decision making score prediction

Abstract

Recommendation systems are widely used to help users acquire interesting information. Most current recommendation systems merely use the overall rating information (Single-Criteria) to recommend items. Some researchers have recently begun to exploit various aspects of an item’s features to more precisely capture the users’ preferences. The technique is called multi-criteria rating. The multi-criteria ratings are usually used to construct the user profiles. However, current multi-criteria recommendation systems still have difficulty updating a user profile depending on time. This report proposes a new multi-criteria rating method that can update user profiles in a required amount of time on an individual basis, and obtain more effective user profiles by exploiting both the user’s preference and behavior profiles. Moreover, to increase the accuracy, we apply multi criteria decision making (MCDM) to the multi-criteria ratings to calculate an item’s prediction value. We conducted experiments under varying conditions using a reliable database, Yahoo movies. The experimental results show that the proposed method outperforms a set of previous methods.