Optimizing Flavonol Content Prediction in Chickpeas: A Comparative Study of Machine Learning Algorithms with NIR Spectroscopy

Abstract

Flavonols in chickpeas offer potential health benefits, necessitating accurate content prediction for nutritional assessment and quality control. This study aimed to develop a rapid, non-destructive method using near-infrared (NIR) spectroscopy and machine learning to predict flavonol concentration in chickpea flour. NIR spectral data from 237 chickpea germplasm accessions were preprocessed and analyzed using four machine learning algorithms: Artificial Neural Networks, Random Forests, Support Vector Regression (SVR), and Decision Tree Regression. The dataset was split into calibration (80%) and validation (20%) sets. The SVR model outperformed others, achieving an RMSE of 0.014 and R² of 0.990 on the calibration set, and an RMSE of 0.086 and R² of 0.853 on the validation set. These results demonstrate the potential of NIR spectroscopy combined with machine learning for rapid and accurate prediction of flavonol content in chickpea flour, supporting efficient screening of germplasm collections.