Comparative Analysis of SVM, k-NN and Logistic Regression Methods in Classifying Turkish Music Genres

Music genre classification represents a fundamental challenge within the field of Music Information Retrieval (MIR). The analysis of audio signals plays a pivotal role in the process of music genre classification, facilitating the extraction of pertinent information from the frequency-based data of the auditory content. In this study, diverse acoustic characteristics were derived through the utilization of the librosa library, and subsequent classification procedures were executed employing machine learning algorithms. For the purpose of this study, a dataset comprising a total of 600 music files in WAV format was meticulously curated. This dataset encompassed six distinct genres, all rooted in Turkish musical traditions. Subsequently, classification tasks were undertaken using Support Vector Machines (SVM), k-Nearest Neighbors (k-NN), and Logistic Regression algorithms. A series of experiments was conducted, varying the kernel functions and distance metrics employed. The findings of this investigation reveal the highest achieved accuracy rates, which amounted to 71.88% with k-NN, 73.44% with Logistic Regression, and 78.65% with the SVM algorithm. Notably, the SVM algorithm demonstrated superior performance in comparison to all other methodologies explored in this study.