An Innovative Approach to Predict Drinking Water Risks Using System, Community, and Regulatory Characteristics.

Area of Science:

• Environmental Science
• Public Health
• Data Science

Background:

• Public drinking water systems (PWS) pose significant health risks and costs.
• Robust predictive models are needed for risk prevention and mitigation.
• Current models often rely on water quality and hydrological data.

Purpose of the Study:

• To compare Logit, Support Vector Machine (SVM), and Extreme Gradient Boosting (XGBoost) models for predicting PWS risks.
• To evaluate the influence of PWS characteristics, community attributes, and regulatory developments on risk prediction.
• To introduce a novel approach focusing on social determinants rather than water quality parameters.

Main Methods:

• Comparative analysis of Logit, SVM, and XGBoost machine learning models.
• Risk prediction based on public drinking water system characteristics, community attributes, and regulatory developments.
• Evaluation of model performance, particularly for health-based risks.

Main Results:

• Extreme Gradient Boosting (XGBoost) demonstrated superior performance compared to Logit and SVM models.
• Community and regulatory characteristics were more influential in risk predictions than PWS characteristics.
• All models showed lower effectiveness in predicting health-based risks.

Conclusions:

• XGBoost offers a cost-effective and suitable alternative for long-term PWS risk forecasting.
• The novel approach using community and regulatory data provides valuable insights for risk anticipation.
• Focusing on social determinants enhances the ability of stakeholders to address PWS risks effectively.