<![CDATA[The classification of music mood characteristics is a crucial instrument in Music Information Retrieval (MIR) systems to support recommendation technology and AI-based emotion analysis. This study aims to evaluate and compare the performance of three classification algorithms: Decision Tree, K-Nearest Neighbors (KNN), and Naive Bayes. The dataset utilized is sourced from the Kaggle Emotion Dataset, comprising 1,440 audio files. The feature extraction process was conducted using the Librosa library to capture acoustic parameters, including Mel-Frequency Cepstral Coefficients (MFCC), Delta-MFCC, Chroma, Spectral Contrast, Spectral Centroid, Spectral Bandwidth, and Tempo. All features were normalized using StandardScaler and distributed into training and testing sets with an 80:20 ratio. Based on the experimental results, the K-Nearest Neighbors algorithm demonstrated the most superior performance with an accuracy of 71.52%. Meanwhile, the Decision Tree algorithm achieved an accuracy of 54.16%, and Naive Bayes obtained 53.47%. The primary contribution of this research is the empirical evidence of the effectiveness of distance-based algorithms in identifying emotional patterns within multidimensional audio data. These findings provide a robust methodological reference for the future development of music emotion recognition systems]]>
Copyrights © 2026
