Supporting data, code and figures for "SHAP-informed tree-ensemble modelling for monitoring-style risk classification of deep excavations"

This archive contains supporting materials for the manuscript entitled “SHAP-informed tree-ensemble modelling for monitoring-style risk classification of deep excavations”. The archive includes Python scripts, summary results, repeated stratified cross-validation outputs, imbalance-treatment ablation results, AUC/PR-AUC summaries, SHAP diagnostic outputs, feature-ranking files, final figures, figure captions, figure alt text, a data dictionary and a source note for the original Kaggle dataset. The original dataset used in the study is the publicly available “Deep Excavation Risk Monitoring Dataset” hosted on Kaggle. The original Kaggle data file is not redistributed in this archive. Users should download it directly from Kaggle and comply with the dataset’s original licence and terms of use. Because the Kaggle repository does not provide full provenance metadata, the dataset is treated in the manuscript as an open monitoring-style benchmark rather than as a verified field-monitoring database. No new field-monitoring data were generated by the authors.