Classification of Arabica Coffee Beans Based on Starters Type with Honey Processing Using Multi-channel Spectral Sensor

Coffee is one of important agricultural commodities in Indonesia, contributing as an income source for farmers and a major export revenue. The specialty coffee industry has begun to utilize microorganisms (starter) in the fermentation process, including the honey process, to obtain a distinctive flavour. However, the use of various starters in this process produces coffee bean with similar color, making it difficult to determine the authenticity of the type of starter used. This research aims to classify arabica coffee beans processed with different types of starters using multi-channel spectral sensor to ensure product quality and authenticity. This research used arabica coffee beans, in the form of green beans, processed with three types of starters, namely Saccharomyces cerevisiae, Lactobacillus sp, and Rhizopus oryzae. Multi-channel spectral sensor was used to acquire the spectra data of coffee sample processed with different starters. The data was then analysed using multivariate analysis based on Partial Least Square – Discriminant Analysis (PLS-DA). In the calibration stage, PLS-DA model built using de-trending pre-treatment was able to predict the type of starter very well, with accuracy, sensitivity, specificity, and precision values, reaching 97%, 95%, 96%, 95%, respectively. This result is also confirmed during validation stage where the built PLS-DA model could predict the type of starter with accuracy, sensitivity, specificity, and precision values, reaching 100%.