This project explores whether injury severity in traffic crashes can be predicted based on environmental and crash-related conditions like weather, lighting, and vehicle damage. Using a publicly available dataset from the Maryland Open Crash Reporting System, we build and evaluate machine learning models to classify injury outcomes.
- Predict Injury Severity: Build ML models to classify injury severity in car crashes.
- Identify Risk Factors: Analyze how different conditions (e.g., weather, driver behavior) influence injury risk.
- Potential Applications:
- Help inform safer driving practices.
- Provide early injury likelihood estimates for concerned family members.
- Explore data-driven approaches for crash risk assessment.
- Language: Python (Jupyter Notebook)
- Libraries:
pandas,matplotlib,seaborn,scikit-learn - Data Source: Crash Reporting – Drivers Data
├── Injury-Classification.ipynb # Main notebook
├── requirements.txt # Project imports
└── README.md # Project overview
- Extensive EDA and visualization of crash and injury data
- Feature cleaning and reduction (handling categorical imbalance and missing values)
- One-hot encoding and scaling
- Supervised learning models for multi-class injury prediction
- (Optional) Unsupervised learning exploration
While predictive models can support driver safety and awareness, they should not replace expert crash investigations or medical evaluations. Model outputs are probabilistic and should be used cautiously.
- Dalia Cabrera
- Ahmed Torki
- Sergio Zavala