Fast authentication of Indonesian ground-roasted Arabica coffee adulterated with roasted soybean by portable LED-based fluorescence spectroscopy and chemometrics analysis.

Abstract

The authentic Arabica coffee renders it vulnerable to fraud and adulteration. Arabica, Robusta, and Liberica green beans appear different. After roasting and grinding, authentic coffee seems identical to inferior varieties. A proper analytical technique for ground-roasted coffee authentication is needed. This study uses portable LED-based fluorescence spectroscopy and chemometrics to identify and quantify roasted soybeans in ground-roasted Arabica. Supervised classification algorithms PLS-DA, LDA, PCA-LDA, and SVMC are compared. Three easy-to-use regression approaches were employed to predict soybean adulteration in adulterated Arabica coffee: PLSR, PCR, and MLR. Classification accuracy was 100 % for the linear kernel-SVMC model, outperforming PLS-DA, LDA, and PCA-LDA. PLSR had the lowest LOD of 4.96 % (w/w) and the best regression model for soybean adulteration (RMSEC of 2.01 % (w/w), RMSECV of 2.10 % (w/w), R2c of 0.99, and R2cv of 0.99). The RPD and RER were 11.39 and 31.07, beyond practical applicability. The proposed method is simpler, non-destructive, and cost-effective authentication method.