Quality Levels Classification of Indonesian Black Tea using Electronic Nose Sensory System Coupled with Ensemble Learning Method
Kombo Kombo (a), Shidiq Nur Hidayat (ab), Trisna Julian (b), Wahyono (c), Kuwat Triyana (a*)

(a) Department of Physics, Universitas Gadjah Mada, Sekip Utara BLS 21, 55281 Yogyakarta, Indonesia
*triyana[at]ugm.ac.id
(b) PT Nanosense Instrument Indonesia, Umbulharjo, 55167 Yogyakarta, Indonesia
(c) Department of Computer Science and Electronics, Universitas Gadjah Mada, Sekip Utara
BLS 21, 55281 Yogyakarta, Indonesia

Abstract

Aroma is considered to be the most important attribute for tea quality determination. In this work, an ensemble learning classifier was proposed to classify the quality levels of Indonesian black tea using an electronic nose sensory system. Four machine learning techniques, namely quadratic discriminant analysis (QDA), extreme gradient boost (XGBoost), k-nearest neighbor (k-NN), and support vector machine with radial basis function (SVMRBF) kernel are used as base classifiers in the ensemble classifier. All incorporated base classifiers and the final outcome of the ensemble model are being compared against several performance metrics namely accuracy, area under the curve score, recall, precision, and F1-score for the classification of black tea quality levels. The experimental results showed that the ensemble model achieve the best performance with the training and testing accuracy results of 100% and 98% respectively which is a noticeable improvement compared to the basic learners. The overall results demonstrated that E-nose sensory system combined with the ensemble learning-based method can improve the efficiency of discriminating black tea quality levels.

Keywords: Aroma, Ensemble learning classifier, Accuracy, Area under the curve, Recall, Precision

Topic: Signal Processing and Communication