This project focuses on Green AI applied to environmental risk modeling. It aims to simulate the displacement of an oil spill on the sea surface and to evaluate machine learning models for prediction and analysis.
The project combines:
- physical simulation,
- data preprocessing,
- machine learning experimentation,
- evaluation of model performance.
The objective is to build efficient and interpretable models while considering computational cost and environmental impact.
Oil spill dispersion is influenced by multiple factors such as:
- currents,
- wind,
- diffusion phenomena.
Accurate simulation and prediction are critical for:
- environmental risk assessment,
- emergency response,
- mitigation strategies.
- Simulation of oil spill displacement over time
- Generation of spatio-temporal data
- Controlled scenarios for experimentation
- Cleaning and structuring simulation outputs
- Feature engineering
- Preparation of datasets for ML models
- Baseline models
- Comparative evaluation of different approaches
- Focus on prediction accuracy and stability
- Quantitative metrics
- Model comparison
- Analysis of trade-offs between performance and complexity
- Python
- NumPy, Pandas
- Scikit-learn
- Simulation models
- Jupyter Notebooks
This project emphasizes:
- efficient use of computational resources,
- model simplicity when possible,
- responsible application of AI to environmental challenges.
This work can be extended toward:
- real-world environmental data,
- more advanced physical models,
- production-ready pipelines and MLOps practices.