ATR-FTIR Coupled With Machine Learning Provides a Fast Method for Identifying and Distinguishing 55 Varieties of...

Area of Science:

• Analytical Chemistry
• Computational Chemistry
• Pharmacognosy

Background:

• Fruit-derived medicinal materials (FDMM) are widely used but suffer from variety confusion, impacting consumer safety and rights.
• Accurate identification of FDMM is essential for quality control, efficacy, and safety assurance.

Purpose of the Study:

• To develop a method for differentiating and identifying 55 kinds of FDMM.
• To combine attenuated total reflection-Fourier transform infrared spectroscopy (ATR-FTIR) with machine learning (ML) techniques for this purpose.

Main Methods:

• Collected 861 sample batches of FDMM for model development and validation.
• Constructed 10 classification models, including K-nearest neighbor (KNN), support vector machine (SVM), and neural networks.
• Evaluated models based on accuracy, precision, recall, and F1-score, selecting the optimal model for validation.

Main Results:

• The K-nearest neighbor (KNN) model demonstrated superior performance, with all evaluation metrics exceeding 0.98.
• The KNN model achieved a prediction accuracy of 85.7% on an independent set of 140 samples.
• The combined ATR-FTIR and ML approach proved effective in distinguishing the majority of FDMM.

Conclusions:

• Attenuated total reflection-Fourier transform infrared spectroscopy (ATR-FTIR) coupled with machine learning (ML) enables rapid and accurate identification of FDMM.
• This methodology contributes to ensuring the quality and safety of fruit-derived medicinal materials.
• The developed models offer a robust solution for addressing variety confusion in the FDMM market.