Related Experiment Video
Updated: Apr 30, 2026
08:47
Author Spotlight: UAV Remote Sensing for Efficient Invasive Plant Biomass Estimation
Published on: February 9, 2024
2.0K
Rapid identification of different processing degrees of Eucommiae cortex based on machine learning and explainable
Yatong Xin1, Weihao Zhu1, Dong Zhou1
1School of Pharmacy, Nanjing University of Chinese Medicine, Jiangsu, Nanjing 210023, China; Jiangsu Key Laboratory of Chinese Medicine Processing, Jiangsu, Nanjing 210023, China.
Food Chemistry
|April 28, 2026
Summary
Machine learning and explainable AI accurately classify Eucommiae cortex processing degrees using physical characteristics. Decoction piece features, particularly breaking elongation, were key indicators for quality control in industrial production.
Area of Science:
- Pharmacognosy
- Computational Chemistry
- Materials Science
Background:
- Eucommiae cortex (EC) processing significantly impacts its quality and therapeutic efficacy.
- Objective classification of EC processing degrees is crucial for pharmaceutical quality control.
- Existing methods for assessing EC processing are often subjective or time-consuming.
Purpose of the Study:
- To develop an intelligent, rapid, and accurate method for classifying four processing degrees of Eucommiae cortex (EC).
- To integrate machine learning (ML) and explainable artificial intelligence (XAI) for objective quality assessment.
- To identify key physical features indicative of EC processing levels.
Main Methods:
- Collected powder characteristics (L*, a*, b*) and decoction piece characteristics (R, G, B, GRAY, percentage of breaking elongation - PBE) using spectrophotometer, high-resolution camera, and material testing machine.
- Evaluated various machine learning models, including XGBoost, for classification accuracy.
- Utilized SHAP (SHapley Additive exPlanations) for feature importance analysis and the Analytic Hierarchy Process (AHP)-Entropy Weight Method (EWM) to analyze appearance-component relationships.
Main Results:
- XGBoost model achieved perfect training accuracy (100%) and high test accuracy (94.74%) in classifying EC processing degrees.
- Explainable AI (XAI) identified percentage of breaking elongation (PBE), a* color value, and 'h' as the most significant features.
- Models utilizing decoction piece characteristics demonstrated superior performance compared to those using only powder features.
Conclusions:
- The integrated ML and XAI approach provides an effective, objective, and rapid method for determining Eucommiae cortex processing degrees.
- Decoction piece characteristics, especially PBE, are critical indicators for quality control.
- This study offers a valuable reference for at-line sampling and quality control in the industrial production of Eucommiae cortex.
