Data and code for: Interpretable machine learning for predicting the eccentric compressive capacity of circular and square CFST columns

This deposit contains the experimental database and Python code supporting the manuscript "Interpretable machine learning for predicting the eccentric compressive capacity of circular and square CFST columns" (Li et al., 2026). Files:database.xlsx: 504 specimens of circular (C-CFST) and square (S-CFST) concrete-filled steel tubular columns under eccentric loading. Includes geometric dimensions, material properties, load eccentricities, and measured ultimate capacities.code.py: Complete Python script for data preprocessing, training six tree-based ensemble models (RF, ERT, GBDT, XGBoost, LightGBM, CatBoost) with Optuna hyperparameter optimization, SHAP interpretability analysis, single-parameter and two-parameter contour analyses, and Monte Carlo uncertainty quantification (±5% input noise). Requirements: Python 3.9+, pandas, numpy, scikit-learn, xgboost, lightgbm, catboost, optuna, shap, openpyxl (for reading .xlsx). Instructions: Run `code.py` directly. The script will load `database.xlsx`, perform the full analysis, and generate the results presented in the manuscript. License: CC BY 4.0Citation: Please cite the associated manuscript and this dataset when using the data or code.