Reproducibility Package for Physically Consistent and Unseen-Mix-Validated Machine Learning for Concrete Compressive Strength Prediction

This archive contains the reproducibility package supporting the manuscript “Physically Consistent and Unseen-Mix-Validated Machine Learning for Concrete Compressive Strength Prediction”. The package includes the experiment scripts, UCI dataset metadata, cleaned data, cross-validation outputs, unseen-mix group validation results, nested tuning outputs, conformal prediction diagnostics, regime-transfer stress tests, SHAP and permutation-importance outputs, figures, tables, manuscript files and validation reports. The package supports reproducibility and audit of the reported applied engineering machine-learning experiments.