Hi, I’m Ajay 👋
This repository contains my project “An Aggregation Approach to Short-Term Traffic Flow Prediction”, where I explored how different traffic conditions contribute to urban noise pollution and built predictive models to estimate noise levels.
Urban noise pollution is one of the most disruptive challenges in modern cities. In this project, I set out to:
- Analyze real-world traffic data (vehicle types, road setups, time-of-day effects).
- Identify key factors contributing to noise levels.
- Compare multiple machine learning models for predictive accuracy.
- Deploy the most effective model for potential use in smart city applications.
- Python
- NumPy, Pandas – data preprocessing
- Matplotlib, Seaborn – visualization
- Scikit-learn – baseline ML models
- XGBoost – final deployed model
I tested a wide range of algorithms, from simple regressions to ensemble methods and neural networks.
| Model | R² Score |
|---|---|
| Gradient Boosting | 0.9989 |
| XGBoost (Final) | 0.9986 |
| Random Forest | 0.9985 |
| Linear Regression | 0.944 |
| Neural Network | -101.24 |
- XGBoost was chosen for its balance of accuracy, speed, and scalability.
- Neural networks surprisingly overfitted and failed, reminding me that complexity isn’t always better.
- Heavy vehicles and narrow lanes significantly increase noise levels.
- Morning and evening rush hours show distinct noise patterns.
- Some features (e.g., Leq, Vehicle Count) had much stronger influence than others.
- Simple metrics like traffic density (vehicles per meter) proved to be powerful predictors.
- Fine-tune hyperparameters of XGBoost.
- Build a web demo to visualize predictions interactively.
- Incorporate additional features like weather conditions and road surfaces for better accuracy.
# Clone the repository
git clone https://github.com/ajay9704/traffic-flow-prediction.git
# Navigate into the project
cd traffic-flow-prediction
# Install dependencies
pip install -r requirements.txt
# Run the model
python main.py- LinkedIn: linkedin.com/in/budida-ajay
- GitHub: github.com/ajay9704
✨ Thanks for checking out my project! If you find it interesting, feel free to ⭐ this repo.