Detecting Music Genre Using Extreme Gradient Boosting

This paper summarizes our contribution to the CrowdAI music genre classification challenge "Learning to Recognise Musical Genre from Audio on the Web'' as part of the WebConference 2018. We utilize different approaches from the field of music analysis to predict the music genre of given mp3 music files, including a convolutional neural network for spectrogram classification, deep neural networks and ensemble methods using various numerical audio features. Our best results were obtained by an extreme gradient boosting classifier.