Music Classification Using Machine Learning Methods

Music widely exists in our daily life. To recognize the music category effectively, the topic namely music classification is to classify the music into various types, such as blues, jazz, pop, etc. With the development of machine learning, music classification using machine learning method has dominated with satisfying performance and real-time speed. In this paper, we aim to analyze the performance of several machine learning methods, including decision tree, random forest, logistic regression and the popular convolution neural network. We firstly preprocess the dataset and convert the music into spectrogram, which is sent to each model. We evaluate those models with mean and standard deviation of correlation rate. We conclude that the convolution neural network obtains the best result, with 75.3% mean correlation rate. This paper provides a brief analysis on music classification using machine learning methods, which can further influence the algorithm design.