Predicting Functionality of Tanzanian Waterpoints

Abstract

In modern Tanzania, many residents rely on local water pumps as their primary source of fresh water. As such, it is important that these pumps remain functional, and that any potential issues are identified and dealt with swiftly for the well-being of the Tanzanian people. Using data collected by the Tanzania Ministry of Water and organized with the Taarifa data management interface, our goal was to accurately predict the function status of waterpoints in Tanzania. We did this using machine learning algorithms to analyze the forty given features to create a model that can assess whether a pump is functional, nonfunctional, or functional but in need of repair. We completed this research as part of a competition hosted by DrivenData, and so we assessed our results using DrivenData's prediction rate accuracy metric.