SHAP and LIME Enhanced Tsunami Detection with Stacking Based Integration of Advanced Tabular Deep Learning Models
8th International Congress on Human-Computer Interaction, Optimization and Robotic Applications, ICHORA 2026, Ankara, Turkey, 21 - 23 May 2026, (Full Text)
- Publication Type: Conference Paper / Full Text
- Doi Number: 10.1109/ichora69329.2026.11537047
- City: Ankara
- Country: Turkey
- Keywords: AutoInt, DANet, Deep Learning, Explainable Artificial Intelligence (XAI), FTTransformer, LIME, NODE, SHAP, Stacking, TabNet, Tsunami Detection
- Hacettepe University Affiliated: Yes
Abstract
This study proposes a stacking based ensemble learning approach for tsunami detection. The model consists of advanced tabular deep learning methods, including TabNet, NODE, FT-Transformer, AutoInt, and DANet. The proposed approach aims to produce promising classification performance and generalization capability by leveraging the complementary learning advantages of different architectures. In addition to stacking model and individual base model performance analysis and comparison, the stacking model's decision-making process is also analyzed by using SHAP and LIME based explainable artificial intelligence techniques to interpret feature contributions. The results indicate that the proposed approach provides both promising performance and explainability, offering an effective and reliable solution for tsunami detection tasks.