Machine Learning Predicts Coal Workers' Pneumoconiosis
Researchers from Chinese medical and engineering institutions (published 2026) trained six machine-learning models on occupational history, lung-function tests, and routine blood markers to predict coal workers’ pneumoconiosis (CWP). Optuna-tuned LightGBM and CatBoost achieved test AUCs of 0.974 and 0.975, while XGBoost reached recall 0.926 and F1 0.952; SHAP highlighted age, FEV1/FVC, and platelet count as top predictors, and performance remained strong without job-type data.
Key Points
- 1Demonstrated high predictive accuracy: LightGBM and CatBoost AUCs 0.974–0.975; XGBoost recall 0.926.
- 2Used routine occupational, lung-function, and blood indicators to enable low-cost early CWP screening.
- 3Enable implementable interpretable models; age, FEV1/FVC, platelet count guide targeted monitoring and interventions.
Scoring Rationale
High-quality, peer-reviewed ML evaluation with strong metrics and interpretability, limited by moderate sample size and single-center retrospective data.
Sources
Public references used for this report.
Practice with real Health & Insurance data
90 SQL & Python problems · 15 industry datasets
250 free problems · No credit card
See all Health & Insurance problems