Music Recommendation System using Collaborative Filtering with SVD

Abstract

This research provides a music recommendation system that creates tailored recommendations for users based on their listening history using a collaborative filtering algorithm and Singular Value Decomposition (SVD). Initially, the research methodology attempted to use cosine similarity to generate recommendations, but it was found to be ineffective due to the inability to handle sparse matrices for large datasets. Therefore, the research shifted its approach to using SVD to overcome this issue. The Amazon Digital Music dataset is used for the implementation of the system, which contains user ratings and reviews for various music products. The dataset is divided into three matrices using the SVD algorithm: the user matrix, the song matrix, and the diagonal matrix. With the use of these matrices, it is possible to forecast missing ratings for unrated products. The predicted ratings are then used to generate personalized recommendations for the user. The Root Mean Square Error (RMSE) and the Mean Absolute Error (MAE) metrics are used to gauge the system's performance. According to the evaluation's findings, the system performs admirably in terms of accuracy and efficacy, with low RMSE and MAE values. This indicates that the system can generate accurate recommendations for users based on their listening history, which can enhance the user experience and engagement with music streaming services. In conclusion, the work highlights the effectiveness of the collaborative filtering algorithm with SVD in generating personalized music recommendations for users. The failure of the initial approach using cosine similarity due to the inability to handle sparse matrices for large datasets underscores the importance of selecting appropriate algorithms for specific datasets. The proposed system demonstrates the effectiveness of using SVD for generating accurate and personalized recommendations for users, and future work could explore other machine learning techniques to further improve the system's performance.