Related Experiment Video
Updated: Feb 28, 2026
12:08
Field Identification of Matricaria chamomilla using a Portable qPCR System
Published on: October 10, 2020
7.4K
Origin Identification of Scutellariae radix Based on Multidimensional Quality Indicators and Machine Learning
Xiao-Lu Liu1, Tong Zhu1,2, Ming-Yue Zhang2
1School of Chinese Materia Medica, Chongqing University of Chinese Medicine, Bishan, Chongqing 402760, China.
Molecules (Basel, Switzerland)
|February 27, 2026
Summary
This study developed a machine learning method to trace Scutellariae radix origins using quality indicators. Random Forest models accurately identified Scutellariae radix origins, outperforming neural networks with limited data.
Area of Science:
- Pharmacognosy
- Computational Chemistry
- Medicinal Plant Research
Background:
- Scutellariae radix is a vital traditional Chinese medicine.
- Accurate origin identification is crucial for quality control and therapeutic efficacy.
- Existing methods for Scutellariae radix origin identification face challenges in accuracy and efficiency.
Purpose of the Study:
- To establish an origin identification method for Scutellariae radix using multidimensional quality indicators and machine learning.
- To enable accurate and rapid traceability of Scutellariae radix from different production areas.
- To compare the performance of various machine learning models for origin identification.
Main Methods:
- Collected 43 batches of Scutellariae radix samples from four origins.
- Measured 12 key quality indicators: flavonoids, physicochemical parameters, chromaticity, and anti-inflammatory activity.
- Constructed and compared five machine learning models: Random Forest (RF), Extreme Learning Machine (ELM), Backpropagation Neural Network (BP), and Radial Basis Function Neural Network (RBF).
Main Results:
- Significant differences (p < 0.05) were observed in key quality indicators among Scutellariae radix from different origins.
- Random Forest (RF) achieved the highest test accuracy (75%) and consistent performance metrics (79.17% precision, recall, F1-score).
- Neural networks (ELM, BP, RBF) showed lower test accuracy (66.67%), with varying performance.
Conclusions:
- Random Forest (RF) is recommended for Scutellariae radix origin identification, especially in small-sample scenarios.
- Ensemble methods like RF effectively mitigate overfitting and enhance generalization compared to neural networks with limited data.
- Future research should focus on expanding data and hyperparameter tuning to further improve classification accuracy.
