Novel Approaches to Improve Collaborative Filtering Recommender System
Abstract
Internet has an ocean of information and hence, number of choices are available which result in information overload. It becomes hard for people to make choice. This issue can be resolved by use of recommender system. Recommender system analyze user data and provide recommendations tothe user. Recommender systems are categorized as content-based, collaborative filtering based, and hybrid recommender system. This paper compares different categories of recommendation system. The main focus of this paper is to highlight limitations of collaborative filtering approach and approaches proposed to overcome the drawbacks of collaborative filtering. Machine learning approaches are also mentioned to differentiate traditional recommender system and machine learning based recommender system. The most populardatasets used for recommendation i.e. Movielens and Epinions datasets statistics are provided in this paper.
Keywords: Recommender system, Collaborative Filtering, Cosine Similarity, Pearson Correlation, MAE, RMSE.