leejay-glitch/tanzania-water-well-prediction

# 💧 Predicting the Functional Status of Tanzanian Water Wells ## 📌 Overview Access to clean and functioning water wells is crucial in Tanzania, where many rural communities depend on them for daily life. This project aims to use machine learning to predict the **functional status** of water wells—whether they are **functional**, **non-functional**, or **need repair**—based on characteristics such as location, installation details, and water quality. This classification task provides practical value to organizations that manage and maintain wells by helping prioritize maintenance and funding for at-risk water points. --- ## 🧠 Business and Data Understanding ### 👥 Stakeholder Audience Our primary stakeholders include: - **Non-Governmental Organizations (NGOs)** working on water access and rural development - **Government agencies** responsible for infrastructure and water resource management - **Local authorities** tasked with well monitoring and community welfare These groups benefit from knowing **which wells are likely to fail** or are already non-functional, enabling **efficient resource allocation**, **proactive repair planning**, and **community health interventions**. ### 🧾 Dataset Description The dataset is sourced from **Tanzania’s Water Point Mapping** initiative, provided by Taarifa and WaterAid. It contains over 59,000 records of water points with attributes such as: - `location`: region, district, ward - `technical details`: installer, construction year, pump type - `management`: funder, scheme management, water quality - `status_group` (target): **Functional**, **Needs Repair**, **Non-Functional** This diverse, real-world dataset is ideal for applying supervised machine learning to predict the status of water wells and gain actionable insights. --- ## 🧪 Modeling Approach We framed the problem as a **multi-class classification** task and experimented with various models, including: - **Logistic Regression** - **Decision Tree Classifier** …