[16a-K505-1~9] 23.1 Joint Session N "Informatics"

10:45 AM - 11:00 AM JST(1:45 AM - 2:00 AM UTC)

[16a-K505-7] Prediction of band gaps from chemical compositions using interpretable machine learning

〇Taichi Masuda¹, Katsuaki Tanabe¹ (1.Kyoto Univ.)

Keywords:

semiconductor,band gap,machine learning

The band gap is a critical physical property that determines the optical and electronic characteristics of semiconductor materials. In recent years, research on band gap prediction using data-driven machine learning approaches has advanced to achieve fast and highly accurate predictions. However, complex models with high predictive accuracy often face challenges in the interpretability of their results. Therefore, this study analyzed machine learning methods for predicting band gaps from chemical compositions by leveraging explainable artificial intelligence techniques. Using explainable artificial intelligence techniques, we identified key physical quantities in band gap prediction models and clarified the dependence of band gaps on the properties of constituent elements. Additionally, we showed that feature selection based on these techniques enhances the predictive accuracy of machine learning models.