Prediction of Bandgap in Lithium-Ion Battery Materials Based on Explainable Boosting Machine Learning Techniques

The bandgap is a critical factor influencing the energy density of batteries and a key physical quantity that determines the semiconducting behavior of materials. To further improve the prediction accuracy of the bandgap in silicon oxide lithium-ion battery materials, a boosting machine learning model was established to predict the material's bandgap. The optimal model, AdaBoost, was selected, and the SHapley Additive exPlanations (SHAP) method was used to quantitatively analyze the importance of different input features in relation to the model's prediction accuracy. It was found that AdaBoost performed exceptionally well in terms of prediction accuracy, ranking as the best among five predictive models. Using the SHAP method to interpret the AdaBoost model, it was discovered that there is a significant positive correlation between the energy of the conduction band minimum (cbm) of silicon oxides and the bandgap, with the bandgap size showing an increasing trend as the cbm rises. Additionally, the study revealed a strong negative correlation between the Fermi level of silicon oxides and the bandgap, with the bandgap expanding as the Fermi level decreases. This research demonstrates that boosting-type machine learning models perform superiorly in predicting the bandgap of silicon oxide materials.