A machine learning-quantitative microbial risk assessment (ML-QMRA) framework for predicting potential health risks from pathogens in drinking water sources

Pathogens in drinking water sources are diverse and pose potential health risks, yet rapid monitoring and risk assessment remain challenging. This study established detection methods for three indicator bacteria and six pathogens, and compared six machine learning models to develop a machine learning-quantitative microbial risk assessment (ML-QMRA) framework for assessing the potential health risks of microbial contamination. The framework provided a feasible data-driven approach for predicting the potential health risks posed by pathogens using routine water quality indicators. Water samples were collected monthly from May 2024 to December 2025, with pathogen concentrations typically ranging from 103 to 106 copies/L. Correlation analysis revealed weak associations between pathogens and indicator bacteria (p > 0.05). Among the six machine learning models, Random Forest and Decision Tree showed better performance. Overall, the predictive models achieved good agreement with the observed data for all pathogens (R2 > 0.75). Notably, the Decision Tree model demonstrated excellent predictive performance for Pseudomonas aeruginosa (P. aeruginosa, R2 > 0.90). SHapley Additive exPlanations (SHAP) analysis was applied to interpret the influence of variables on model predictions, revealing that models for indicator bacteria were primarily influenced by water temperature and turbidity, while TDS and electrical conductivity were the most influential predictors of adenovirus. The findings indicate that conventional water quality parameters show useful predictive potential for estimating pathogen levels and associated potential health risks.