Performance evaluation of music genre prediction system using convolutional neural network and random forest.

To evaluate the performance of the Music Genre Prediction System using a novel Convolutional Neural Network and Random Forest Algorithm. Materials and Methods: The dataset used in this study is the standard dataset called GTZAN, which contains 1000 music files, each 100 of 10 different genres. Spectrograms and Mel-frequency cepstral coefficients (MFCC) are derived from the music files for training the model. In both algorithms,20 iterations are done with different sample sizes. The pretest power obtained is 0.08. Comparative analysis of mean accuracies is made between two algorithms using SPSS software. Result: From the experimental data analysis it is observed that the Convolutional Neural Network gives an accuracy of 79.23% while the Random Forest gives an accuracy of 64.67%. Conclusion: This study shows that the novel Convolutional Neural Network performs better when compared to the Random Forest for predicting the music genre accurately. [ABSTRACT FROM AUTHOR]