Spectral-based Features Ranking for Gamelan Instruments Identification using Filter Techniques

Abstract

Abstrak Kata kunci: support vector machine, transkripsi otomatis, Gain Ratio, ekstraksi fitur Abstract In this paper, we describe an approach of spectral-based features ranking for Javanese gamelan instruments identification using filter techniques. The model extracted spectral-based features set of the signal using Short Time Fourier Transform (STFT). The rank of the features was determined using the five algorithms; namely ReliefF, Chi-Squared, Information Gain, Gain Ratio, and Symmetric Uncertainty. Then, we tested the ranked features by cross validation using Support Vector Machine (SVM). The experiment showed that Gain Ratio algorithm gave the best result, it yielded accuracy of 98.93%.

Improved performance of fake account classifiers with percentage overlap features selection

“…To gain deeper insights into these features' characteristics, we employ a ranking technique founded on information gain (IG), a widely recognized algorithm in machine learning research for generating feature rankings. IG quantifies feature rankings based on the system's entropy [18], [19]. Additionally, we introduce a quantitative measure known as the percentage overlap (PO) to evaluate a feature's capacity to distinguish between different classes.…”