EFFECT OF NORMALIZATION OF GENRE MUSIC DATA ON CLASSIFICATION PERFORMANCE WITH RANDOM FOREST

Abstract

This research is about the classification of the music genre using the Random Forest method. This test uses a dataset from GitHub or GITZAN about the music genre with 10 labels, 26 features and 1000 total data. This research is divided into two stages, namely by classifying all data without being normalized, and by using all normalized data. . In this research, Min-Max is used for data normalization method, and for accuracy calculation using Confusion Matrix method. The resulting accuracy when using all data with data that is not normalized produces an accuracy of 66.3%, while the resulting accuracy performance when using all data with normalized data results in an accuracy of 65.1%.