Using Data Mining to Understand Drinking Water Advisories in Small Water Systems: a Case Study of Ontario First Nations Drinking Water Supplies

Abstract

Although access to safe drinking water is widely assumed to be universal, small drinking water systems in many countries continue to experience an unacceptably large number of drinking water advisories (DWAs). The goal of this research is to describe novel data mining tools that identify the factors contributing to DWAs in small drinking water systems. A dataset containing information related to First Nations drinking water systems in the Province of Ontario, Canada is used for the case study. A decision tree classifier (one of the fastest and most versatile predictive modeling algorithms currently available for data mining) visually maps out the relationship of system characteristics (e.g., source water, system age, and operator certification) to DWA likelihood. The developed model achieves an overall accuracy of 71 % during repeated cross-validation of predictive performance and is of utility when prioritizing future expenditures aimed at proactively reducing the risk of delivering compromised water.