Bandgap prediction of MXene materials via multiple machine learning approaches

Multiple machine learning methods were employed to evaluate their effectiveness in predicting the bandgap of MXene materials. Three representative models—GBDT, LightGBM, and SVR—were systematically compared using performance metrics including MAE, RMSE, and R 2 . The results indicate that the GBDT model achieved the highest prediction accuracy, particularly excelling in MAE and R 2 . Further feature importance analysis revealed that the presence of a bandgap (IsGap) is the most critical factor influencing prediction, significantly outweighing other features. Additionally, structural and electronic descriptors such as lattice constant (LC) and valence band maximum (VBM) were also found to contribute to bandgap determination. These findings suggest that tuning semiconductor-related properties is an effective strategy for bandgap control in MXene materials. Overall, this work provides a robust, low-cost, and efficient machine learning-based framework to facilitate high-throughput screening and rational design of MXene materials.