Hybrid based Collaborative Filtering with Temporal Dynamics

Abstract

Hybrid-based collaborative filters use some part or entire database relating to user preferences for making recommendations for new products and new users. In our time, it is of utmost importance to make recommendations in line with interests and demands of users by making their interest alive. However, although Hybrid-based collaborative filters are used in this area, changing of preferences of users in a time, emergence of new products and new users overshadow success of such systems. Traditional hybrid-based collaborative filtering (CF) technique become insufficient for responding interests and demands changing in a time. For this reason, temporal changes in recommendation systems become an important concept. Together with the study conducted, an appropriate and new method has been developed in line with changing pleasure and demands depending on time. In the recommended system, unlike traditional hybrid technique based CF technique, point given to the products depending on dates scored by users has been attempted to be estimated. In this study, process has been made over netflix data for measuring success of both traditional hybrid based CF technique and the recommended system. Quite successful and rewarding results have been obtained in the issue of accuracy of predicted points. Keywords- Recommendation System;, Data Mining; Temporal Dynamics.