Hitness of Original Pilipino Music can be predicted up to 92%

Abstract

In this work, musical features (e.g. pitch, tempo, and dynamic range) are extracted from Original Pilipino Music (OPM) released from 2004 to 2006 and are used to forecast hit songs. We show that a trained feed-forward NN can predict which songs would hit up to 92.3% regardless of musical genre. The accuracy is remarkable given that linear classifers are just within the proportion chance criterion.