Geographical origin discrimination of Chenpi using machine learning and enhanced mid-level data fusion

Area of Science:

• Food Science
• Analytical Chemistry
• Computational Chemistry

Background:

• Chenpi, or dried tangerine peel, is a traditional Chinese ingredient with recognized medicinal and culinary value.
• The geographical origin of Chenpi significantly influences its quality, active compounds, and market value.
• Accurate identification of Chenpi's origin is crucial for quality control and authentication.

Purpose of the Study:

• To develop and validate a strategy for distinguishing Chenpi samples based on their geographical origin.
• To evaluate the effectiveness of machine learning methods combined with data fusion for Chenpi origin classification.

Main Methods:

• Analysis of 39 Chenpi samples from eight regions in Xinhui district using gas chromatography (GC) and mid-infrared (MIR) spectroscopy.
• Application of four machine learning algorithms (K-nearest neighbors, artificial neural network, etc.) for discrimination.
• Implementation of two mid-level data fusion strategies to integrate GC and MIR data.

Main Results:

• Data fusion strategies significantly enhanced the accuracy of Chenpi origin discrimination.
• Modified mid-level data fusion combined with K-nearest neighbors and artificial neural network models achieved the highest performance.
• The best models misclassified only one sample, demonstrating high classification accuracy.

Conclusions:

• Machine learning approaches, particularly when combined with modified mid-level data fusion, offer an effective method for classifying Chenpi samples by geographical origin.
• This strategy provides a robust tool for authenticating Chenpi and ensuring product quality.
• The findings support the use of analytical techniques and computational methods in traditional ingredient authentication.