Salrahim21/Tanzanian-Water-Well-Analysis

Pump it Up: Data Mining the Water Table Can you predict which water pumps are faulty to promote access to clean, potable water across Tanzania? # 📘 Project: Binary Classification of Water Pump Functionality This project applies machine learning techniques to classify the functionality status of water pumps in Tanzania. The dataset is converted into a binary classification problem: predicting whether a water pump is **functional (1)** or **non-functional/needs repair (0)**. --- ## 📁 Files - `Notebook_Annotated.ipynb` — Jupyter Notebook containing the full workflow: preprocessing, resampling, training, tuning, and evaluation. --- ## 🔧 Problem Setup - **Original Labels:** - `functional` - `functional needs repair` - `non functional` - **Binary Mapping:** - `functional` → 1 (Positive Class) - `functional needs repair`, `non functional` → 0 (Negative Class) --- ## ✅ Methodology 1. **Preprocessing** - Convert sparse matrix features to DataFrame - Map target labels to binary 2. **Resampling** - Downsample the majority class (`functional`) - Upsample the minority class (`non-functional/needs repair`) using `resample` from `sklearn.utils` 3. **Modeling** - Train a `DecisionTreeClassifier` with class balancing - Tune hyperparameters using `GridSearchCV` with stratified 5-fold cross-validation - Train a `RandomForestClassifier` with grid search 4. **Evaluation** - Performance is evaluated on test data using: - **Classification Report:** Precision, Recall, F1-score - **Confusion Matrix** - Both macro and weighted averages reported --- ## 🔍 Example Results (Random Forest) | Metric | Class 0 (Functional) | Class 1 (Non-Functional) | |------------|----------------------|---------------------------| | Precision | 0.80 | 0.75 | | Recall | 0.86 | 0.66 | | F1-score | 0.83 | 0.70 | | Accuracy | 78% | | --- ## 📈 Best Parameters Found (Random Forest) ``` { 'max_depth': None, 'min_samples_leaf': 1, 'min_samples_split': 5, 'n_estimators': 100 } ``` …