Rapid differentiation of Atractylodes chinensis and Atractylodes japonica using FT-IR spectroscopy, HPLC combined with Machine Learning

Atractylodes japonica is frequently confused with or used as a substitute for Atractylodes chinensis in commercial products due to their morphological similarities, posing risks to product quality and safety. This study developed a rapid, multi-platform approach to differentiate these two species using Fourier Transform infrared (FT-IR) spectroscopy, high-performance liquid chromatography (HPLC), and machine learning. First, FT-IR data (81 spectra × 967 variables) were processed using multiple preprocessing techniques and the random frog algorithm, which selected 87 key spectral features. Based on these features, machine learning models including Extreme Learning Machine (ELM), Support Vector Machine (SVM), and Random Forest (RF) achieved 100 % classification accuracy. The robustness of this approach was further validated by correctly classifying an external data set of 54 unknown mixed samples. Furthermore, HPLC fingerprinting combined with chemometrics identified five differential chemical markers. Models built using only these five markers also achieved 100 % classification accuracy. This integrated FT-IR and HPLC strategy offers a robust and interpretable solution for authenticating Atractylodes chinensis , significantly improving quality control and preventing adulteration in functional foods and herbal products. • Integrated FT-IR spectroscopy and HPLC identifies Atractylodes species. • Machine learning models achieved 100 % classification accuracy. • Five chemical markers were identified for precise quality control. • External validation confirmed robustness on unknown mixed samples. • Prevents adulteration in functional foods and herbal products.