Penetapan Kadar Kafein pada Sampel Bubuk Kopi dengan Menggunakan Metode Nir-Kemometrik

Abstract

This study aims to determine the caffeine content in ground coffee samples from three major species Arabica, Robusta, and Liberica using Near-Infrared (NIR) Spectroscopy combined with chemometric approaches. The optimal calibration model was obtained using the Partial Least Squares (PLS) method based on the original spectrum without transformation, yielding a coefficient of determination (R²) of 0.960 and a Root Mean Square Error (RMSE) of 0.057 during calibration. Validation was performed using Leave-One-Out Cross Validation (LOOCV) and external testing with a test set, demonstrating stable and accurate model performance. Caffeine quantification in real samples using the NIR method showed no statistically significant difference (p > 0.05) compared to the reference method, KLTKT-Densitometry, for both the original PLS model and the smoothed spectrum. These findings confirm that the NIR-chemometric approach can serve as a reliable, non-destructive, efficient, and accurate alternative for coffee quality analysis, and offers potential for species classification based on spectral characteristics