Detection and classification of indonesian civet and non-civet coffee based on statistical analysis comparison using E-Nose

Abstract

Civet coffee is a highly priced premium beverage in Indonesia. Because of its high economic value, civet coffee is often falsified with non-civet coffee. The detection and classification of coffee aroma using an e-nose has been the subject of several researches. However, only few researches have been done on civet coffee and non-civet coffee detection using an e-nose. This study aimed to improve the classification between civet coffee and non-civet coffee by trying out different combinations of classification methods and statistical parameters. The coffee aroma data were taken from e-nose sensors with different sensitivity toward certain chemicals. There are a number of steps in the classification of coffee aroma: ground truth data acquisition, statistical feature extraction, classification, and performance evaluation. The experimental results of this study indicate that an e-nose can recognize and distinguish well between civet and non-civet coffee. Comparing 6 classes of coffee, the best performing combination was the decision tree algorithm with the average and standard deviation parameters, which obtained 97% accuracy.