Movie Recommendation using Multiple Data Sources

Most of the recommender systems are built for the content or item providers. For example, Netflix recommends movies or TV shows, Amazon recommends books or other items being sold on Amazon, Facebook or Twitter recommends popular posts or tweets, Pinterest recommends related pins, YouTube recommends videos, etc. Most of these recommender systems are designed based on the usage data collected on their own websites.

However, sometimes it could be helpful if we could get information about recommended items from multiple data sources, providing multiple perspectives for users to make their decisions. In this research work, we study different approaches to integrate multiple data sources and the effect on the recommendation results when multiple data sources are used to recommend items. We propose a multiple data source-based movie recommender system that uses MovieLens rating data, YouTube movie trailer data, Netflix rating data, and tweets from Twitter. The user feedback data such as likes, dislikes, comments on movie trailers posted on YouTube can be helpful side information for movie recommender systems. In this research work, we study the effect of adding these side information to the movie rating data. Our proposed recommendation framework can integrate the trailer and rating data adopting various integration strategies: integrating all the trailer data as movie features, using sentiment scores derived from the trailer comments as a rating matrix to integrate with the movie rating matrix, and treating others as the movie features, or only integrating the sentiment score based rating matrix with the movie rating matrix. Our experiment shows that if we include the movie trailer data, recommendation accuracy is improved. We also find that the most accurate result is achieved if all the trailer feedback data is integrated as movie features. We use both Matrix Factorization (MF) and Deep Neural Network (DNN) Models to design our system. We find that the DNN model performs better than the MF model. Our results show that when we include multiple data sources to recommend items using the DNN model, the recommendation accuracy (F1 score) is increased by 41% on average comparing to the case when only one data source is used.