Abstract
Music genre categorization has been considered to be an essential task within the context of music data recovery. Genres serve as categories or labels that enable the classification of music based on shared attributes, including musical style, instrumentation, cultural origins, historical context, and other distinctive elements. The purpose of classifying music genres is to automatically assign music pieces to one or more predefined genres. The present research suggests a new method for music genre identification via integrating deep learning models with a metaheuristic algorithm. The proposed model uses a pre-trained Zeiler and Fergus Network (ZFNet) to extract high-level features from audio signals, while an Extreme Learning Machines (ELM) is utilized for efficient classification. Furthermore, the model incorporates a newly developed metaheuristic algorithm called the Modified Electric Eel Foraging Optimization (MEEFO) algorithm to optimize the ELM parameters and enhance overall performance. To evaluate the effectiveness of the model, it has been tested on two widely recognized benchmark datasets, namely GTZAN and Ballroom, and the results are contrasted with some advanced models, comprising MusicRecNet, Parallel Recurrent Convolutional Neural Network (PRCNN), RNN-LSTM, ResNet-50, VGG-16, Deep Neural Network (DNN). The outcomes demonstrated that the suggested system surpassed several existing methods regarding precision, recall, and accuracy.