Based on near-infrared spectroscopy and chemometrics to rapidly evaluate the adulteration of Ganoderma lingzhi powder

Abstract Ganoderma lingzhi, the dry fruiting bodies of G. lucidum or G. sinensis, is a microbial food of high nutritional and health value. It is expensive but in high demand. In pursuit of higher profits, counterfeit products adulterated with G. lingzhi, such as G. applanatum, have appeared in the market. To identify the authenticity and forecast the degree of adulteration in Ganoderma lingzhi powder rapidly and non-destructively, the combination of near-infrared spectroscopy (NIRS) and chemometrics was used. Partial least squares discriminant analysis (PLS-DA), back propagation neural network (BPNN), support vector machine (SVM), and random forest (RF) were adopted as qualitative identification of G. lingzhi authenticity model methods, and partial least-squares (PLS) was developed as a quantitative prediction of adulteration content. Preprocessing and feature variables selection methods were developed to optimize the model and screen the best model. Among these experimental approaches, PLS-DA + first-order derivatives (D1), SVM + D1 + Competitive adaptive reweighted sampling (CARS), RF + standard normal variate transform (SNV) and BPNN + D1 + Uninformative variable elimination (UVE) + CARS achieved 100% classification accuracy. SVM + second-order derivatives (D2) + CARS and BPNN + D2 + CARS identified all adulterated G. lucidum, PLS-DA + D1 + UVE + CARS, RF + D2 + Genetic algorithm (GA), SVM + D2 + GA, and BPNN + D2 + CARS could distinguish all adulterated G. sinensis effectively.