This paper presents a novel approach for music genre classification using deep learning methodologies. The method involves preprocessing input signals and representing their characteristics using a combination of Mel Frequency Cepstral Coefficients (MFCC) and Short-Time Fourier Transform (STFT) features. Two Convolutional Neural Networks (CNNs) are applied to process these features, with each model's hyperparameters adjusted using the Black Hole Optimization (BHO) algorithm to minimize training error. The combined features from the CNNs are then classified using a SoftMax classifier. The effectiveness of the proposed method is evaluated using the GIZAN and Extended-Ballroom datasets, achieving classification accuracies of 95.2% and 95.7%, respectively, demonstrating superior performance over previous methods. The BHO algorithm's role in optimizing CNN hyperparameters and enhancing classification accuracy is highlighted, making the proposed approach a promising solution for music genre classification.This paper presents a novel approach for music genre classification using deep learning methodologies. The method involves preprocessing input signals and representing their characteristics using a combination of Mel Frequency Cepstral Coefficients (MFCC) and Short-Time Fourier Transform (STFT) features. Two Convolutional Neural Networks (CNNs) are applied to process these features, with each model's hyperparameters adjusted using the Black Hole Optimization (BHO) algorithm to minimize training error. The combined features from the CNNs are then classified using a SoftMax classifier. The effectiveness of the proposed method is evaluated using the GIZAN and Extended-Ballroom datasets, achieving classification accuracies of 95.2% and 95.7%, respectively, demonstrating superior performance over previous methods. The BHO algorithm's role in optimizing CNN hyperparameters and enhancing classification accuracy is highlighted, making the proposed approach a promising solution for music genre classification.