Orandifelix/PumpInsight

This project builds a machine learning model to predict the condition of water wells across Tanzania using data from the Pump it Up: Data Mining the Water Table competition # 💧 PumpInsight: Predicting Water Well Functionality in Tanzania > This is a machine learning classification project built on real-world water infrastructure > data from Tanzania helping NGOs and government teams move from reactive to > proactive well maintenance. --- ## Table of Contents - Overview - Business and Data Understanding - The Problem - Stakeholders - Dataset - Key EDA Findings - Modeling - Data Preparation - Models Tested - Evaluation - Final Model Results - Per-Class Performance - Feature Importance - Business Implications - Conclusion - Repository Structure - How to Reproduce - Links and Sources --- ## Overview **PumpInsight** is a supervised machine learning project that predicts the operational status of water wells across Tanzania using a three-class classification model: | Class | Meaning | | ------------------------- | --------------------------------------------- | | `functional` | Well is operational and serving the community | | `functional needs repair` | Well is working but requires maintenance | | `non functional` | Well has completely failed | The project covers the full data science pipeline — business understanding, data exploration, preprocessing, iterative modeling, and final evaluation — and is built on the DrivenData Pump it Up competition dataset. | | | | ----------------- | ---------------------------------------- | | **Final Model** | Tuned Random Forest (`n_estimators=200`) | | **Test Accuracy** | 80.3% | | **Test Macro F1** | 0.685 | | **Problem Type** | Multi-class Classification | | **Dataset Size** | 59,400 labeled wells × 40 features | --- ## Business and Data Understanding ### The Problem Tanzania, a Sub-Saharan African nation of over 57 million people, face …