Research on Performance of the Classifying Models Based on Chinese, Pakistani, and Other Genres

In recent years, with the increase in the amount of audio on the internet, the demand for audio classification is increasing. This paper focuses on finding the performance of the classifiers, uses Python for the simulation part, compares the performance, and finds the best classifier. Two experiments are performed for this paper; for the first part of the experiment, Pakistan and Chinese music samples are considered, and classifiers are used to classify these music samples. It is found that the artificial neural network (ANN) has lowest accuracy of 81.4%; additionally, support vector machine (SVM), k-nearest neighbor (KNN), and convolutional (CNN) accuracies remain between 82% to 86% based on the dataset. Random forest model has the highest accuracy of 94.3%. It is considered to be the best classifier. For the second part of the experiment, other genres such as classical, country, and pop music were added to the previous dataset. After adding these genres, performance of the classifying models varies slightly; it fluctuates between 75% to 84%. These results can be used for music recommendation applications.