Qualitative analysis of terpenoid esters based on near-infrared spectroscopy and machine learning.

Abstract

This study delves into the method of qualitative analysis of terpenoid esters using near-infrared spectroscopy technology. Terpenoid esters are bioactive compounds widely used in the pharmaceutical and cosmetics industries. Near-infrared spectroscopy technology enables rapid and accurate component analysis without compromising the integrity of the sample, which is particularly important for valuable samples that need to be preserved intact or require subsequent analysis. This research combines machine learning techniques, such as K-Nearest Neighbors (K-NN) classifier, Random Forests algorithm, and Back Propagation Neural Networks (BPNN), to analyze terpenoid ester samples extracted from different concentrations of eluents, and compares and evaluates these algorithms. This study results show that in the test set, the prediction accuracy of the K-NN classifier is 96.154% and BPNN is 94.231%, and the Random Forest algorithm performs the best with a prediction accuracy of 100%. Additionally, this study utilizes the Random Forest algorithm to predict the characteristic spectra of terpenoid esters, demonstrating the effectiveness of feature spectrum extraction by ensuring a prediction accuracy of 100% while reducing the number of spectral features.