Original experimental data for the manuscript “Machine learning-based source term joint probability region inversion for uncertainty quantification in plutonium explosion accidents”

This dataset contains the original experimental data supporting the manuscript “Machine learning-based source term joint probability region inversion for uncertainty quantification in plutonium explosion accidents”. It includes the original JPSTI simulation dataset, held-out test-set results, joint Monte Carlo sample arrays, Optuna hyperparameter-optimization records, cross-validation records, and source data for manuscript Figures 2-5. The package also includes file-level metadata, checksums, a data dictionary, a dataset inventory, an NPZ array index, and English/Chinese file-structure guides for reviewer verification. The dataset supports the final comparison between JPSTI and three retained direct joint baselines: MultiTargetXGBoostLSS_MVN, MultiTargetXGBoostLSS_MVT, and MultivariateNGBoost_MVN. The JPSTI independent joint baseline, trained model artifacts, and source code are not included in this review data package.