Item recommending by item co-purchasing network and user preference

Abstract

A novel recommendation approach based on multiple-objectives optimization is introduced. It utilizes histories of item co-purchases as represented by directed graphs and histories of user preferences for items as represented by user-item rating matrices to create models of user behavior. Herein, multiple objectives for item recommendations are considering to maximize the ratings and rankings of each item as derived from user preferences and co-purchasing networks of items, respectively. Finally, optimal items, called Pareto optimal solutions, will be recommended.