Bartok: Music Time Period Classification

Previous work has high classification accuracy when classifying music genres that are very different [1] (e.g. rock vs. pop, jazz vs. classical), but little machine learning research has been done to classify music by subgenres within a larger genre that describe temporally and stylistically similar music. In this paper, we apply machine learning to classify classical music by time period. To do so, we used supervised learning algorithms (Naïve Bayes and SVM), unsupervised algorithms (knearest neighbors), and ensemble algorithms (AdaBoost) on a uniformly distributed data set, comprised of over 4,000 MIDI files extracted from an online collection. We then analyze and discuss the performance of our classifier.