A Hybrid Approach to Developing a Movie Recommender System

Abstract

As information is growing rapidly, which makes users to locate the information they need ineffectively. Therefore, Recommender Systems (RS) have been developed in the past 10 years in order to help users find information more quickly and that can also help companies to understand customers’ behaviors for doing next step promotion activities. Most of the existing RS may not be good at producing accurate recommendations to users because they have not taken any contextual information when giving recommendations (e.g. Users Context, Item Context, Time, etc.). Unlike Context-Aware Recommender System (CARS) which is able to take extra contextual information to further analyze and describe users’ actual situation to return more meaningful recommendations [1]. Context-Aware Movie Recommender System is implemented by using Collaborative Filtering (CF) framework that is integrated with factorization machine approach as illustrated that is a generic model. CARS is making use of contextual information (e.g. Location, Emotion, Mood) to provide more relevant recommendations to users. In offline experiment, it has been compared with other approaches (e.g. SVD++, KNN Item-Based) that show the accuracy of recommendations has significant improvement.