Machine Learning Models to Classify Shiitake Mushrooms (Lentinula edodes) According to Their Geographical Origin...

Area of Science:

• Agricultural Science
• Computational Biology
• Food Science

Background:

• Shiitake mushroom (Lentinula edodes) consumption has surged globally, ranking second worldwide.
• Consumers increasingly demand transparency regarding the geographical origin of shiitake mushrooms due to nutritional and health benefits.
• Distinguishing the origin of shiitake is crucial for market integrity and consumer trust.

Purpose of the Study:

• To develop and evaluate machine learning algorithms for determining the geographical origin of shiitake mushrooms consumed in Korea.
• To assess the predictive accuracy of models based on stable isotope data (δ13C, δ15N, δ18O, δ34S).

Main Methods:

• Utilized literature-reported experimental data including stable isotope ratios and origin information.
• Developed and validated machine learning models: Random Forest and Support Vector Machine.
• Tested models on datasets categorized into two and three origin groups (Korean, Chinese, Chinese inoculated sawdust blocks).

Main Results:

• Random Forest achieved high accuracy (0.940) and kappa (0.908) for three origin categories.
• Support Vector Machine demonstrated superior performance (accuracy 0.988, kappa 0.975) for two categories (Korean vs. Chinese including sawdust blocks).
• Random Forest also excelled (accuracy 0.952) for a different two-category classification (Korean vs. Chinese excluding sawdust blocks), with acceptable testing phase accuracies (0.839-0.964).

Conclusions:

• Machine learning algorithms are effective tools for identifying the geographical origin of shiitake mushrooms.
• The developed models show strong predictive capacity suitable for real-world applications.
• This research supports enhanced traceability and authenticity verification in the shiitake mushroom market.