Music Recommendation System Design Based on Gaussian Mixture Model

Abstract

The paper establishes a double-layer classifier based on Gaussian Mixture Model and the Thayer model to divide the music style into several categories. On the basis of effective verification of experiments, the music listening experience is added into the model to analyze and normalize two-dimensional data points and the tastes of music users or playing times also can be added to obtain new three-dimensional data points. Then the Gaussian Mixture Model is employed again for classifying the new three-dimensional data points. In this way, not only can the taste changing process of users for different music be analyzed, but also the similarity among different users can be calculated. Therefore, music can be recommended properly to music users.