Classification of Music Genres based on Machine Learning SVM and CNN

This study aims to compare the performance of two classification methods—Support Vector Machine (SVM) and Convolutional Neural Network (CNN)—in identifying music genres based on audio data from the GTZAN dataset. Data in a 30-second audio file is converted into a numerical and visual representation (Mel Spectrogram) for each model. The evaluation used precision, recall, and f1-score metrics and was supported by a confusion matrix. Results showed that SVM had higher accuracy (75%) than CNN (71.5%), with more stable performance in most genres. CNN showed superiority in certain genres, such as classical and hip-hop, but experienced a decline in accuracy in genres such as disco and rock. This analysis confirms that the selection of classification models should consider the characteristics of data and genres and open up opportunities for hybrid approaches or the use of more complex architectures in music classification.