Quantitative Analysis of Sorghum Starch With Machine Learning and Near-Infrared Spectroscopy

A model based on NIR technology and machine learning algorithm is established for rapid quantification of sorghum starch. Using brewing sorghum as the primary raw material, and we collected near-infrared diffuse reflectance spectra using a Fourier transform near-infrared spectrometer, and the isolated forest algorithm is used to eliminate abnormal samples, optimize the dataset, then establish the quantitative model of sorghum starch by using multiple modeling software to compare multiple preprocessing methods and regression models. After Isolation Forest preprocessing, the Partial Least Squares Regression (PLSR) model achieved optimal performance (R _c ²=0.993, RMSECV = 3.401), demonstrating high efficiency and accuracy for rapid starch quantification. The model is efficient, rapid, and accurate, and is suitable for the quantitative analysis of sorghum starch, which provides technical support for the quality control of raw materials for brewing and the accurate procurement of brewing materials.