Addressing gaps in data on drinking water quality through data integration and machine learning: evidence from Ethiopia

This study addresses the scarcity of reliable drinking water quality data in Ethiopia by integrating existing household survey data with machine learning techniques. Predictive models were developed using data from the 2016 Ethiopia Socio-Economic Survey (ESS) to identify households with contaminated drinking water (≥1 E. coli per 100 mL). The best-performing model (Random Forest) achieved high accuracy (88.5%) and discrimination (AUC 0.91), effectively predicting water quality based on demographic, socioeconomic, and geospatial variables. This model was successfully applied to other ESS waves lacking water quality testing, demonstrating its potential for filling data gaps in drinking water safety monitoring.

Alemayehu A. Ambel, Robert Bain, Tefera Bekele Degefu, Ayca Donmez, Richard Johnston, Tom Slamyaker