A novel approach for music recommendation system using matrix factorization technique

Abstract

A recommender system provides personalized content to its users to handle the ever expanding information overload, thus improving customer relationship management. Music is a subject used widely across the world in different aspects of life. Music recommender systems help users to listen to the right choice of music. With the advent of mobile devices and Internet, access to different music resources is easily available. In this paper, we provide music recommendations to the Million Song Dataset using the TuriCreate’s core ML library and with a focus on two methods of collaborative filtering techniques: user-based and item-based recommendations. Results are deduced exploring numerous metrics to measure the similarity of users and items such as cosine metric, Pearson correlation, latent matrix factorization and others. A comparison of different evaluation metrics is carried out to check for the effectiveness of the recommender system.