Optimized Convolutional Neural Network Using Genetic Algorithm for Music Genre Classification
Hana Magdelina Yumil, Florence Sia, Tan Soo Fun, Lai Po Hung · 2024
Music genre classification plays an important role in identifying and classifying music tracks into genre. Due to the enormous number of genres available nowadays, music genre classification has been automated through analyzing the features of audio files of music by using machine learning specifically the Convolutional Neural Network (CNN) method. However, the hyperparameter values of CNN has not yet been optimized to ensure accurate classification. It has been found that the classification effectiveness of CNN can be further enhanced with the right setting of hyperparameter in many different domains. Hence, this paper proposes to optimize the hyperparameter setting of CNN by using Genetic Algorithm (GA) for music genre classification. The hyperparameter includes learning rate, batch size, and number of epoch. Experiments have been conducted to assess the effectiveness of the optimized CNN algorithm by comparing it to a baseline CNN method on music genre dataset. The results show that the optimized CNN can classify music genre well.