Classification of Popular Music Genre Using Convolutional Neural Network Method with Data Augmentation

Music genre is a character of the music itself. The development of the music industry today has given birth to various kinds of music genres. The continuous evolution of the music industry has given rise to an array of genres, contributing to a rich tapestry of musical expression. Classification of a music genre often causes debate because of differences in experience or perception in assessing music, because in classifying music there are various parameters such as the instrument, how to play, and the tone. The study’s objective is to advance music genre classification by synergizing the CNN model with data augmentation, aiming for heightened accuracy and performance. The research employs K-fold cross-validation, distributing a 9:1 ratio for training and validation data. The optimal model, incorporating data augmentation, a 25% dropout rate, and segmenting songs into five divisions, achieves a notable validation accuracy of 88.8%. This marks a significant stride in enhancing the precision of music genre classification algorithms, contributing to the evolving landscape of music analysis and recommendation systems.