Performance Evaluation of Chemometric Prediction Models—Key Components of Wheat Grain

Abstract

The present study was aimed to evaluate the accuracy of using near-infrared spectroscopy (NIRS) for predicting protein, moisture, starch and ash content values of wheat. The physiochemical properties of wheat were predicted using twelve prediction models of preprocessing coupled with regression tools. The performance measure of SVM aided with extended multiplicative scatter correction gave confident prediction results of protein, moisture, ash and starch content with R² values of 0.989, 0.987, 0.976, 0.998 and RMSECV values of 0.263, 0.285793, 0.369 and 0.03 respectively. These results indicate the practical applicability of NIRS in wheat grain quality profiling.