Data Mining to Identify Attributes Associated with Drinking Water Advisories in First Nations Communities

AbstractMaintaining acceptable drinking water quality in First Nations communities has been an ongoing issue in Canada, resulting in frequent drinking water advisories (DWAs). Information about drinking water systems on reserves was combined with historical advisory data to investigate the relationships between system attributes and DWAs using data mining. Decision trees were developed to identify key factors linked to the occurrence, frequency, duration, and causes of DWAs. The results showed that trends varied considerably by province, despite federal management. Operator treatment certification was shown to be associated with the duration of advisories, where systems with fully trained operators were less likely to have long-lasting DWAs. The findings from data mining can be useful for informing future regulatory and funding decisions in efforts to reduce DWAs.