Development of machine learning models using multi-source data for geographical traceability and content prediction...

Area of Science:

• Food science and technology
• Pharmacognosy
• Analytical chemistry

Background:

• Growing demand for Eucommia ulmoides leaf products necessitates robust quality and safety assessments.
• Traditional methods for evaluating medicinal plants often lack speed and scientific rigor.
• Geographical traceability and chemical content prediction are crucial for ensuring the quality of Eucommia ulmoides leaves.

Purpose of the Study:

• To develop and evaluate machine learning models for geographical traceability and chemical content prediction of Eucommia ulmoides leaves.
• To investigate the influence of data preprocessing techniques and data fusion strategies on model performance.
• To establish an effective strategy for the rapid and scientific quality evaluation of Eucommia ulmoides leaves.

Main Methods:

• Utilized multi-source data including infrared spectroscopy.
• Applied traditional machine learning algorithms such as partial least squares discriminant analysis (PLS-DA) and partial least squares regression (PLSR).
• Explored various data preprocessing methods (e.g., first derivative) and low-level data fusion strategies.

Main Results:

• Achieved 100% accuracy in geographical traceability using PLS-DA with low-level fusion of two infrared spectroscopy techniques and first derivative preprocessing.
• Optimal PLSR models for aucubin, geniposidic acid, and chlorogenic acid demonstrated satisfactory predictive performance (RPD > 2.0).
• The PLSR model for quercetin showed weaker predictive ability (RPD = 1.541), indicating a need for further refinement.

Conclusions:

• The proposed strategy effectively enables geographical traceability and chemical content prediction for Eucommia ulmoides leaves.
• Low-level data fusion of infrared spectroscopy techniques combined with specific preprocessing significantly enhances traceability accuracy.
• This research offers valuable insights and methodologies for the quality evaluation of other food and medicinal plants.