Classification of Lupinus seeds into sweet and bitter categories using VIS–NIR spectroscopy and machine learning

Josefa Díaz Álvarez^1,2*Francisco A. Galea-Gragera³Francisco Chávez de la O⁴Pedro A. Salguero-López⁴Fernando Llera Cid³

• ¹University of Extremadura, Badajoz, Spain
• ²Universidad de Extremadura. Tecnología de los Computadores y Comunicaciones. Centro Universitario de Mérida, Mérida, Spain
• ³Pasture and Forage Crops Area, Finca La Orden-Valdesequera” Agricultural Research Institute, which belongs to the Extremadura Scientific and Technological Research Centre (CICYTEX),, Guadajira (Badajoz), Spain
• ⁴Dpto. Ingeniería de Sistemas Informáticos y Telemáticos. Centro Universitario de Mérida. Universidad de Extremadura,, Mérida, Spain

ABSTRACT Purpose: The Lupinus germplasm includes sweet and bitter materials distinguished by compounds responsible for bitterness. Conventional identification is often destructive. This study assesses a non-destructive approach based on visible–near infrared (VIS-NIR) spectroscopy and machine learning to classify whole seeds from seven Lupinus species into sweet or bitter classes. Methods: Five machine-learning algorithms were evaluated on two datasets (reflectance and absorbance) acquired with VIS-NIR spectroscopy. Analyses were conducted on raw spectra and on spectra transformed using four spectral-transformation techniques. Because classes were imbalanced, five resampling methods were compared to improve classification performance. Results: Performance was assessed using F1-score and ROC-AUC. On reflectance, LGR and SVC reached 92.5% and 92.0%; on absorbance, SVC and RF achieved 93.2% and 92.5%. Hybrid transformations consistently improved discrimination, and resampling reduced overfitting associated with class imbalance. Conclusion: The results indicate that combining VIS–NIR spectroscopy with machine learning provides a suitable non-destructive alternative to discriminate sweet and bitter Lupinus materials/ecotypes.