Czech J. Food Sci., 2026, 44(5):424-440 | DOI: 10.17221/58/2026-CJFS

SW-NIR spectroscopy and PLSR for predicting cocoa nibs quality across processing stagesOriginal Paper

Riana Listanti ORCID...^1,2, Rudiati Evi Masithoh¹, Arifin Dwi Saputro¹, Hanim Zuhrotul Amanah¹, Sonny Ilham Wicaksono³

¹ Department of Agricultural and Biosystems Engineering, Faculty of Agricultural Technology, Universitas Gadjah Mada, Yogyakarta, Indonesia

² Department of Agricultural Engineering, Faculty of Agriculture, Universitas Jenderal, Purwokerto, Indonesia

³ Department of Physics, Faculty of Mathematics and Natural Sciences, Universitas Gadjah Mada, Yogyakarta, Indonesia

Cocoa beans quality parameters are among other things biochemical properties such as moisture content, reducing sugars, fat, protein and total phenolic content (TPC) changed during postharvest processes. Current analytical approaches to these parameters are usually not only destructive and time-consuming but also laboratory-based analyses. This study aimed at assessing the potential of short-wave near infrared (SW-NIR) spectroscopy (954–1 700 nm) combined with partial least squares regression (PLSR) for the prediction of these quality parameters in both fresh, fermented, dried and roasted Forastero cocoa nibs. In total, a portable Flame-NIR+ spectrometer analysed 400 samples. Before developing the model, we have compared several spectral preprocessing methods like Savitzky–Golay derivatives, multiple scatter correction (MSC), standard normal variate (SNV), and normalisation. The spectra exhibited various absorption features linked to O-H, C-H, and N-H links reflective of evolving compositions across proximate processing stages. The prediction coefficients (R²P) obtained for moisture content (0.984), reducing sugars (0.958), protein (0.992), and total phenolic content (0.954) of the cocoa nibs was good, while fat prediction showed lower predictive performance than the other evaluated parameters. Savitzky-Golay derivative preprocessing consistently improved the performance of models across all evaluated parameters. These results show that SW-NIR spectroscopy with PLSR can be applied for fast, non-destructive and simultaneous estimations of multiple quality attributes in cocoa nibs during different stages of postharvest processing.

Keywords: biochemical composition; chemometrics; model interpretation; non-destructive analysis; spectral preprocessing

Received: April 6, 2026; Revised: August 23, 2026; Accepted: August 25, 2026; Prepublished online: October 23, 2026; Published: October 30, 2026  Show citation