Project Overview

Access to clean water remains a critical challenge in Tanzania. Approximately 25-30 % of water pumps are not-functional, limiting daily water access for millions of people.

This project builds a machine learning–based decision support system to predict water pump operational status, helping stakeholders prioritize maintenance and allocate limited resources more effectively.

Business Problem & Objective

| Problem Statement

Field teams and decision-makers often lack early indicators of pump failure, resulting in delayed maintenance and inefficient resource allocation.

| Objective

To predict the operational status of water pumps into three categories:

• Functional
• Functional needs repair
• Non functional

The model is designed to support preventive maintenance planning rather than reactive repair.

| Evaluation Metric

Macro F1 Score, chosen to:

• Handle class imbalance
• Ensure fair performance across minority failure classes

Data Overview