Morocco Wildfire Predictions
# Wildfire Prediction Model - Morocco
This Python script uses machine learning to predict wildfire risk in Morocco. It trains and evaluates models to understand wildfire patterns and identify important risk factors.
## What the Model Does
* **Predicts Wildfire Likelihood:** Estimates the probability of wildfire occurrence based on environmental and temporal data.
* **Uses Machine Learning:** Employs LightGBM (Gradient Boosting) and Random Forest models, with LightGBM showing strong performance.
* **Identifies Key Risk Factors:** Reveals which factors are most influential in predicting wildfires, such as:
* Location (latitude, longitude, sea distance)
* Vegetation health (NDVI, Soil Moisture)
* Temperature patterns
* **Provides Performance Insights:** Evaluates model accuracy, AUC-ROC, and feature importance to understand model behavior.
* **Test Predictions:** Allows for testing the model on new, example data via `synthetic_wildfire_data.csv`.
## Explore the Findings
After running, look at these outputs:
* **Console Logs:** Examine the validation reports for Random Forest and LightGBM to compare model performance metrics like Accuracy and AUC-ROC. LightGBM should show better results.
* **Feature Importances:** The script outputs the top features for both models. Notice the consistent importance of location, vegetation, and temperature factors. This helps understand what drives wildfire risk in the model.
* **ROC Curve:** A graph visualizing LightGBM's performance (True Positive Rate vs. False Positive Rate). Higher AUC-ROC score (closer to 1.0) indicates better model discrimination.
* **Synthetic Data (Optional):** If you provided `synthetic_wildfire_data.csv`, check the updated file to see predicted wildfire probabilities and classifications for your examples.
## Notes
* **LightGBM is the Stronger Model:** Validation results show LightGBM outperforming Random Forest for this wildfire prediction task.
* **Key Factors Identi …