HuldahCR/Tanzanian_Water_Well_Risk_Model

Predicting rural water well functionality in Tanzania using geospatial and explainable machine learning. # 💧 Tanzanian Water Well Functionality Prediction - Phase Three Data Science Project ## 📁 1. Project Overview This project focuses on analyzing and modeling the functionality of water wells in **Tanzania**. The dataset captures key features related to the **condition, installation, management, and geographical attributes** of wells. The goal is to create predictive models that help identify non-functional or at-risk wells, supporting **government agencies, NGOs, and water resource managers** in better targeting interventions. The project follows the **CRISP-DM methodology** and leverages **predictive analytics** to generate actionable insights that can enhance water accessibility and sustainability across communities. --- ## 🧠 2. Business Understanding - **Client/Context**: Tanzanian government agencies, NGOs, and water sustainability programs - **Objective**: Predict the operational status of water wells to improve maintenance planning, reduce downtime, and allocate resources effectively - **Problem Statement**: A large percentage of wells in Tanzania are non-functional, leading to wasted investments and water scarcity for communities - **Metrics of Success**: - Achieve high **accuracy, recall, and F1-score** in predicting well functionality - Identify the **top contributing factors** affecting well sustainability - Provide clear **recommendations** for prioritizing repairs and installations --- ## 📊 3. Data Understanding - **Source**: Tanzania Water Wells Dataset (Taarifa / DrivenData) - **Description**: - The dataset contains ~59,000 records of water wells across Tanzania - Features include `funder`, `installer`, `gps_height`, `population`, `construction_year`, `water_quality`, `quantity`, `source_type`, `management`, `payment`, `region`, and more - Target variable: `status_group` (functional, non-functional, functional needs repair) --- ## 🧹 4. Data Cleaning and Preparation - Imputed missing values for key categorical and numerical columns - Bucketed …