Design and Application of Music Genre Classification Algorithm Based on Machine Learning

Traditional classification methods often depend on manual annotations and expert judgments, which are time-consuming, labor-intensive, and inefficient for large-scale data. This paper introduces a music genre classification algorithm based on machine learning to enhance efficiency and accuracy. The study begins by preprocessing audio data through format conversion and wavelet threshold denoising. Features are then extracted using short-time Fourier transform (STFT) to obtain MFCC (Mel-Frequency Cepstral Coefficients), along with chroma and rhythm features. For model construction, a deep learning model combining convolutional neural network (CNN) and recurrent neural network (RNN) is employed to capture spatiotemporal dependencies in the audio data. Experiments were conducted using two large-scale datasets with a total scale of 500,000 audio tracks, and the results show that the proposed model achieved high classification accuracy on both datasets, with a test set accuracy of 89.3% on the online music platform dataset. In addition, compared with traditional machine learning models, the algorithm proposed in this study performed better in terms of classification accuracy and overall recall rate. This study not only improves the accuracy of music genre classification but also provides technical support for fields such as music recommendation and copyright management.