This project implements a music genre classification system using handcrafted audio features extracted from raw .wav files.
Features such as MFCCs, chroma, spectral, and rhythm descriptors are used to represent musical characteristics.
Random Forest and XGBoost classifiers are trained and evaluated for multi-class genre prediction.
XGBoost achieved a marginal (~0.5%) accuracy improvement over Random Forest, indicating that both models perform at a comparable level on the extracted features.