[Near-infrared spectrum quantitative analysis model based on principal components selected by elastic net]

Area of Science:

• Chemometrics
• Spectroscopy
• Statistical modeling

Context:

• Quantitative analysis of agricultural products like wheat is crucial for quality control.
• Traditional methods may lack accuracy or efficiency in predicting specific components such as protein content.
• Near-infrared (NIR) spectroscopy offers a rapid, non-destructive method for analyzing agricultural samples.

Purpose:

• To evaluate the effectiveness of the Elastic net regression technique for building quantitative analysis models.
• To assess the prediction accuracy and stability of an Elastic net-based model for wheat protein content determination.
• To compare the performance of Elastic net with other established methods like Principal Component Regression (PCR) and Partial Least Squares (PLS).

Summary:

• Elastic net, incorporating L1 and L2 penalties, was utilized to select spectral principal components from 89 wheat samples.
• A quantitative analysis model was established correlating near-infrared spectra with wheat protein content, achieving a correlation coefficient (R) of 0.9849 and a mean relative error of 2.48%.
• The model demonstrated stability and superior prediction accuracy compared to PCR, and comparable accuracy to PLS, confirming Elastic net's suitability for chemometrics.

Impact:

• Elastic net provides a robust and accurate method for quantitative analysis in agricultural science.
• The study validates the use of Elastic net for variable selection and prediction in spectroscopic analysis.
• This approach can lead to more efficient and reliable quality assessment of wheat and potentially other agricultural commodities.