Rapid and non-destructive origin traceability of various citrus-medicine and food homologous substances using near-infrared spectroscopy and variable screening-SVM algorithms

In this work, a rapid and non-destructive approach for the origin traceability of three citrus-medicine and food homologous substances (MFHs) was established for the first time. The method was based on near-infrared (NIR) spectroscopy and chemometrics with integration time optimization. The spectra of Aurantii Fructus Immaturus , Aurantii Fructus and Exocarpium Citri Grandis from different origins were directly collected using a portable NIR spectrometer at three integration times (25 ms, 45 ms, and 65 ms) without damaging the samples. Origin traceability models were constructed with two variable screening-support vector machine (SVM) algorithms. The performance of these models was evaluated using the validation set and external validation set obtained one month later. The results showed that the variable screening strategy could effectively extract characteristic variables, simplify the SVM models and improve the identification ability. 100% identification accuracies for Aurantii Fructus Immaturus and Exocarpium Citri Grandis , and accuracies greater than 95% for Aurantii Fructus were obtained with Monte Carlo uninformative variable elimination (MCUVE)-SVM. These selected variables indicated that the main differences among citrus-MFHs of different origins were attributed to variations in the variety and content of carbohydrates. Besides, optimization of integration time can improve the origin traceability performance.