A Hybrid Approach for Modern Music Recommendation System using Neural Networks and Feature Level Fusion

Abstract: The custom music recommender supports users' favorite songs, which are stored in a huge music database. To predict only the user's favorite songs, the management of the user's preference information and the genre rating is required. In our study, a very short feature vector obtained from a low-dimensional projection and already developed audio features is used for the music genre classification problem. We apply a metric distance learning algorithm to reduce the dimensionality of the feature vector with little performance degradation. We propose the system through the automatic management of user preferences and gender classification in the personalized music system. This Recommender System uses a feature level fusion to combine multiple perspectives and gives an outcome that suits all types of users. The performance of this system is compared with existing legacy system.