Forest Fire Modeling: A Statistical Mechanics Perspective

This repository contains a spatial-temporal analysis of forest fires in Portugal (the infmous UCI forestfire dataset), with a critical evaluation of standard regression-based modeling. Using data visualization, coordinate-wise recurrence analysis, and insights from statistical mechanics, we argue that fire behavior exhibits state-transition characteristics that invalidate naive regression approaches.

🧠 Core Insight

• Fire area prediction is not a regression problem.
• Each coordinate experiences at most one significant burn → indicating fuel exhaustion.
• Without fuel state data, the system becomes partially observable and non-ergodic.
• This approach bridges environmental modeling, spatial statistics, and stochastic state theory.

📁 Contents

• Portugal_FF.ipynb: Jupyter notebook containing full analysis.
• Forest_Fire_Theory_Insight.pdf: Theoretical write-up.
• chart_spatial_pattern.png: Visualization of fire coordinates by month.
• data/forestfires.csv: Original dataset (UCI Forest Fire Dataset).

▶️ How to Run

1. Clone the repository:

   git clone https://github.com/yourusername/ForestFire_StochasticInsight.git
   cd ForestFire_StochasticInsight

2. Install required libraries (use virtualenv if needed):

   pip install pandas matplotlib seaborn

3. Launch the notebook:

   jupyter notebook Portugal_FF.ipynb

📜 License

This work is shared under the MIT License.
For academic or research reuse, please cite the PDF write-up.

🤝 Contact

For collaborations or inquiries, reach out via GitHub or connect on LinkedIn.