CharityPM/tanzanian-water-wells-prediction
Domaine:
environment and energygeospatialType de record:
datasetmodel
Créateur:
Cha
Hôte:
predicting Tanzania water-well functionality
# 💧 Waterpoint Status Classification in Tanzania
This project focuses on predicting the functionality status of waterpoints in Tanzania using machine learning. Access to clean water is essential, and identifying non-functional waterpoints ensures timely maintenance and resource allocation.
## 🔍 Project Objectives
- **Classify** waterpoints into two categories:
- `functional`
- `needs_attention`
- **Compare** multiple classification models.
- **Select** the best performing model.
- **Evaluate** using relevant metrics like accuracy, precision, recall, and F1-score.
- **Visualize** results for clear interpretation.
---
## 📂 Dataset
The dataset consists of waterpoints across various regions in Tanzania, with features including:
- `amount_tsh` (total static head in meters)
- `gps_height`
- `population`
- `well_age`
- `installer`, `basin`, `region`, `scheme_management`, etc.
Target variable:
- `status_group`: Whether a waterpoint is `functional` or `needs_attention`.
---
## 🧪 Models Developed
- **Logistic Regression**:
Accuracy = 77.49%
- Good at identifying functional wells.
- Missed several wells needing attention.
- **Decision Tree**:
Accuracy = 77.0%
- High recall for functional wells.
- Moderate performance for identifying faulty wells.
- **Random Forest (Baseline)**:
Accuracy = 81.87%
- Balanced performance across both classes.
- Best among base models.
- **Random Forest (Tuned)**:
Accuracy = 82%
- Improved recall and F1-score.
- Robust and reliable classifier.
---
## 📊 Exploratory Data Analysis & Visualization
### Load the data
```python
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
# Load the dataset
df = pd.read_csv('waterpoint_data.csv')
df.head()
Sample Visualizations
python
Copy
Edit
# Class distribution
sns.countplot(data=df, x='status_group')
plt.title('Waterpoint Status Distribution')
plt.xticks(rotation=15)
plt.show()
# Population vs Status
sns.boxplot(data=df, x='status_group', y='population')
plt.yscale('lo …