Using Machine Learning to Predict Classical Composers from Audio

Abstract

This research paper explores the application of machine learning techniques to predict classical composers from audio recordings. Classical music, with its rich history and diverse styles, poses a challenge in identifying composers solely based on musical characteristics. The study utilizes a dataset of 2000 Western classical tracks from different eras and employs artificial neural networks for feature engineering. The goal was to develop an accurate predictive model that lists potential composers for a given piece. The results indicate that the LSTM model achieves moderate accuracy, correctly identifying the true composer within the top three predictions. This research is an important contribution to the field as it will further demonstrate the utility of using machine learning to predict the composer of a piece of classical music. It displays the possibilities of using machine learning for other music-related data. It also contributes to the development of future tools that are helpful for both passive music enjoyers as well as professional musicians.