Residual-Corrected XGBoost for Interpretable Prediction of Oxide Double Perovskite Band Gaps

This record contains the processed dataset and Jupyter Notebook code used in the manuscript “Residual-Corrected XGBoost for Interpretable Prediction of Oxide Double Perovskite Band Gaps”. The dataset is based on the double_perovskites_gap dataset from matminer and includes the target variable gap_gllbsc and Magpie elemental descriptors after removing GSbandgap-related features to avoid information leakage. The accompanying notebooks provide the workflow for data preprocessing, model training, residual compensation, statistical evaluation, and SHAP analysis.