Perbandingan Metode Ensemble Learning Random Forest Dan Adaboost Pada Pengenalan Chord Instrumen Piano Dan Gitar

Abstrak

This paper presents a comprehensive analysis of Random Forest and Adaboost algorithms for classifying music chords, focusing on feature extraction and model optimization. The distribution of original data, including audio signal lengths and class distributions, is examined, revealing consistent characteristics across major and minor datasets. Mel Frequency Cepstral Coefficients (MFCC) are extracted with predefined parameters, ensuring feature extraction consistency. Subsequent feature normalization and oversampling using SMOTE address class imbalances. Evaluation metrics, including accuracy, precision, recall, and F1-score, are employed to assess model performance. The results demonstrate Random Forest's superiority in accuracy, precision, and recall over Adaboost. Furthermore, employing RandomizedSearchCV optimization enhances both models' performance, with Random Forest achieving an accuracy of 0.84 and Adaboost attaining 0.80. Confusion matrices illustrate Random Forest's higher prediction accuracy for both positive and negative classes compared to Adaboost. These findings underscore the effectiveness of Random Forest in accurately classifying music chords and highlight the significance of hyperparameter optimization in enhancing classification model performance