Bandgap analysis of transition-metal dichalcogenide and oxide via machine learning approach

Abstract

Predicting bandgap is a crucial topic in materials informatics, however, it is still difficult when the available dataset is limited and unbalanced. Here, we applied a machine learning approach to construct a prediction model for transition metal dichalcogenides and oxides. Using an oversampling technique and atomistic feature engineering, we successfully constructed the machine learning model and analyzed the correlation with other physical properties. Furthermore, we also utilized the model to obtain a compressive sensing model based on physical quantities for analytic interpretation and quick prediction.