https://github.com/kurkigal/air-quality-ml-comparison
https://github.com/kurkigal/air-quality-ml-comparison
Last synced: about 1 month ago
JSON representation
- Host: GitHub
- URL: https://github.com/kurkigal/air-quality-ml-comparison
- Owner: KurKigal
- License: mit
- Created: 2025-08-26T13:21:17.000Z (about 2 months ago)
- Default Branch: main
- Last Pushed: 2025-08-26T13:27:28.000Z (about 2 months ago)
- Last Synced: 2025-08-26T18:03:30.446Z (about 2 months ago)
- Language: Jupyter Notebook
- Size: 1.15 MB
- Stars: 0
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# Air Quality Model Performance Comparison
This project provides a comprehensive analysis comparing Linear Regression, Decision Tree, and Random Forest algorithms for air quality prediction using temporal pattern analysis.
## 🎯 Project Objective
To compare the performance of different machine learning algorithms for predicting Benzene (C6H6) concentration and analyze temporal patterns in air quality data.
## 📊 Dataset Information
- **Source**: UCI ML Air Quality Dataset
- **Size**: 9,471 observations × 17 features
- **Temporal Range**: Hourly measurements
- **Target Variable**: C6H6(GT) - Benzene concentration
- **Features**: CO, NOx, NO2 levels, temperature, humidity, sensor data
## 🔧 Technologies Used
```python
pandas==1.5.0
numpy==1.23.0
scikit-learn==1.1.0
matplotlib==3.6.0
seaborn==0.12.0
scipy==1.9.0
```
## 🔍 Analysis Performed
### Exploratory Data Analysis (EDA)
- ✅ Data distributions and histogram analysis
- ✅ Correlation matrix calculation
- ✅ Temporal pattern analysis (hourly/daily/weekly)
- ✅ Relationships between pollutants and weather variables
- ✅ Outlier detection and box plot analysis
### Data Preprocessing
- Missing value imputation (median strategy)
- Feature engineering (hour, day, month, day of week)
- StandardScaler normalization
- Time-series aware train/test split (70%-30%)
## 🤖 Models Compared
1. **Linear Regression**: Basic linear model
2. **Decision Tree Regressor**: Non-linear pattern capture
3. **Random Forest Regressor**: Ensemble learning (100 estimators)
## 📊 Model Performance Results
| Model | MSE | RMSE | MAE | R² |
|-------|-----|------|-----|-----|
| **Decision Tree** | 0.0021 | 0.0462 | 0.0115 | **0.999999** |
| **Random Forest** | 0.0032 | 0.0565 | 0.0128 | **0.999999** |
| **Linear Regression** | 15.9253 | 3.9906 | 3.6325 | 0.9949 |
## 🎯 Key Findings
- **Strongest Correlations**:
- C6H6(GT) ↔ AH (Absolute Humidity): r = 0.9846
- C6H6(GT) ↔ Temperature: r = 0.9714
- NOx(GT) ↔ NO2(GT): r = 0.8171
- **Temporal Patterns**: Distinct hourly and weekly cycles in pollutant levels
- **Model Performance**: Tree-based models achieved exceptionally high accuracy (R² ≈ 1.0)
## ⚠️ Analysis Notes
The extremely high R² values (0.999999) may indicate:
- Strong feature-target correlations
- Potential overfitting
- Pattern similarity in time-series data
## 🚀 Installation & Usage
### Requirements
```bash
pip install pandas numpy scikit-learn matplotlib seaborn scipy
```
### Data Download
Download the dataset from Kaggle:
```bash
kaggle datasets download -d fedesoriano/air-quality-data-set
```
### Execution
```bash
jupyter notebook air-quality-model-comparison.ipynb
```
Or if using Jupyter Lab:
```bash
jupyter lab air-quality-model-comparison.ipynb
```
## 📈 Technical Details
- **Data Split**: Time-aware split (temporal order preserved)
- **Feature Engineering**: Time-based features added
- **Scaling**: StandardScaler applied
- **Evaluation**: MSE, RMSE, MAE, R² metrics used
## 📚 Key Insights
1. **Decision Tree Performance**: Achieved the lowest error metrics with near-perfect R² score
2. **Random Forest Robustness**: Slightly higher errors but still excellent performance
3. **Linear Model Limitations**: Higher errors indicate non-linear relationships in data
4. **Feature Importance**: Weather variables (temperature, humidity) are strong predictors
5. **Temporal Dependencies**: Clear patterns in hourly and weekly pollution cycles
## 🔬 Research Applications
This analysis can be valuable for:
- Environmental monitoring systems
- Air quality prediction models
- Urban planning and policy decisions
- Public health assessments
- Industrial emission control
## 👤 Developer
- **Kaggle**: [@kurkigal](https://www.kaggle.com/kurkigal)
## 📄 License
MIT License
## 🤝 Contributing
Contributions are welcome! Please feel free to submit a Pull Request.
---
⭐ If you found this project helpful, please consider giving it a star!