Collaborative Filtering Recommendation using Matrix Factorization: A MapReduce Implementation

Matrix Factorization based Collaborative Filtering (MFCF) has been an efficient method for recommendation. However, recent years have witness the explosive increasing of big data, which contributes to the huge size of users and items in recommender systems. To deal with the efficiency of MFCF recommendation in the context of big data challenge, we propose to leverage MapReduce programming model to re-implement MFCF algorithm. Specifically, we develop a four-step process of MFCF, each of which is implemented as MapReduce tasks. The experiments are conducted on a Hadoop cluster using a real world dataset of Netflix. The empirical results confirm the efficiency of our method.