Related Experiment Video
Updated: Jul 17, 2026
08:03
Heuristic Mining of Hierarchical Genotypes and Accessory Genome Loci in Bacterial Populations
Published on: December 7, 2021
2.9K
Enhancing kelp origin prediction accuracy: A TabPFN model with stable isotope analysis and explainability techniques
Xuming Kang1, Yanfang Zhao1, Zhijun Tan2
1Yellow Sea Fisheries Research Institute, Chinese Academy of Fishery Sciences. Key Laboratory of Testing and Evaluation for Aquatic Product Safety and Quality, Ministry of Agriculture and Rural Affairs, Qingdao 266071, China.
Food Chemistry
|February 26, 2026
Summary
Accurate kelp origin prediction is vital for quality control. An explainable TabPFN model accurately identified kelp origins using stable isotope data, offering significant speedups over traditional methods.
Area of Science:
- Marine Biology
- Food Science
- Computational Chemistry
Background:
- Kelp origin traceability is crucial for ensuring product quality and safety in the food industry.
- Stable isotope analysis provides a potential method for geographical origin determination due to regional variations in environmental factors.
- Traditional methods for kelp origin identification may lack accuracy or efficiency.
Purpose of the Study:
- To develop and validate an accurate and efficient method for identifying the geographical origin of kelp in China.
- To explore the utility of stable isotope ratios (C, N, O, H) in kelp for origin prediction.
- To leverage explainable artificial intelligence for robust traceability.
Main Methods:
- Kelp samples from diverse Chinese origins were analyzed for stable isotope ratios of carbon, nitrogen, oxygen, and hydrogen.
- An explainable TabPFN (Tabular Predictive Factorization Network) model was employed for classification.
- Local Interpretable Model-agnostic Explanations (LIME) was used to interpret model predictions.
Main Results:
- Stable isotope ratios of carbon, nitrogen, and oxygen showed significant regional variations, while hydrogen did not.
- The TabPFN model achieved 100.0% accuracy, precision, recall, and F1 score, with an AUC of 1.000, in just 2.5 seconds.
- The model demonstrated a 5782-fold speedup compared to tuned traditional machine learning models.
- LIME provided insights into the contribution of each isotope to origin prediction.
Conclusions:
- Stable isotope analysis combined with the TabPFN model offers a highly accurate and efficient approach for kelp origin identification in China.
- The explainable nature of the model enhances trust and understanding of the prediction process.
- This methodology provides a robust tool for quality and safety management in the kelp industry.
