Development and comparison of regression models for determination of starch in chickpea using NIR spectroscopy

Crop quality characteristics are rapidly and efficiently assessed using near-infrared spectroscopy. Over the last several decades, NIR spectroscopy’s advent and broad application have been an enormous success story in analytical technology development. NIR spectroscopy is frequently used in agricultural and food goods to identify and quantify an unlimited number of analytes. The near-infrared area has a wavelength range of 800 to 2500 nm. Machine learning approaches have proven to be highly successful at predicting various agricultural crop components. The concentration of the starch component in Chickpea (Cicer arietinum L.) whole-grain flour was determined using NIR spectroscopy data and machine language algorithms. Starch prediction models are developed using Linear Regression (LR), Artificial Neural Network (ANN), Random Forest (RF), Support Vector Regression (SVR), and Decision Tree Regression (DTR) algorithms. Performance of the models is evaluated using measures, namely, Root Mean Square Error (RMSE), Residual Standard Error (RSE), Coefficient of Determination (R²), and Adjusted Coefficient of Determination (adjusted R²). It was observed that LR outperformed all other models in terms of accuracy for predicting starch components from preprocessed spectra, with RMSE, RSE, R² and adjusted R² values of 0.03, 0.04, 0.98, and 0.97, respectively. The accuracy of the ANN model is similar to that of the LR, with minor differences in RMSE, RSE, R² and adjusted R², values of 0.03, 0.04, 0.97, and 0.97, respectively.

• Near-infrared spectroscopy (NIRS) with machine learning algorithms is one of the most advanced non-destructive component prediction assessment techniques available.

• The NIRS technique has been successfully used for the rapid analysis of starch, moisture, protein, and fat content in many agricultural and food products since its first application in the 1960s.