JUDAH004/TANZANIAN-WATER-WELLS-PROJECT

A classification project on the status of wells/ pumps in Tanzania # 💧 CLASSIFICATION PROJECT:TANZANIAN WATER WELLS. ## 📝 Project Description This project tackles a real-world problem from **Tanzania**, where thousands of water pumps are installed across rural regions. The goal is to **predict the operational status** of a water pump (`functional`, `non functional`, or `functional needs repair`) using historical and geographic data. This information is critical for resource allocation, maintenance prioritization, and policy decisions. ## 📊 Dataset Overview - **Features include**: - **Categorical**: installer, basin, scheme_management, etc. - **Numerical**: amount_tsh, gps_height, population, etc. - **Binary**: public_meeting, permit - **Target Variable**: - `functional` - `non functional` - `functional needs repair` --- ## ✅ Project Steps 1. **Data Cleaning** - Missing value treatment - Outlier handling - Type casting 2. **Feature Engineering** - Grouping rare categories - Handling categorical, continuous, and binary features separately - Optional one-hot encoding or ordinal encoding 3. **Modeling** - Baseline: Logistic Regression - Simple: Decision Tree Classifier - Evaluation using precision, recall, F1-score, and ROC AUC 4. **Evaluation & Visualization** - Classification reports - Confusion matrix - ROC Curve (Macro-average for multiclass) --- ## 📈 Results Summary | Metric | Logistic Regression | Decision Tree | |---------------|---------------------|----------------| | Accuracy | 72.3% | **74.8%** | | Precision | 69.7% | **74.9%** | | Recall | 72.3% | **74.8%** | | F1 Score | 69.3% | **74.9%** | | AUC (macro) | ~0.78 | **~0.81** | The **Decision Tree model** outperformed the baseline Logistic Regression in all key metrics. --- ## 📊 Visualizations - 1. Feature importance bar chart - Distribution of target class against numerical features - confusion matrix - distribution of target class - …