MichaelHead20/algerian-forest-fire-classifier

Python, pandas, scikit-learn, seaborn Built logistic regression and decision tree classifiers to predict forest fire occurrence using meteorological data. Logistic regression achieved an AUC of 0.998. # 🔥 Algerian Forest Fire Classifier Predicting forest fire occurrence in Algeria using meteorological data and the Fire Weather Index (FWI) system. Built as part of CS-165 Data Science at Swansea University. --- ## Overview Algeria experiences severe forest fires during the summer months, a problem worsening with climate change. This project uses a dataset of 242 records from two Algerian regions (June–September 2012) to explore whether FWI components — alongside temperature, humidity, wind, and rainfall — can accurately classify fire vs. non-fire conditions. Two classification models were trained and evaluated: | Model | AUC Score | |---|---| | Logistic Regression | 0.998 | | Decision Tree | 0.967 | Both models demonstrate strong predictive accuracy, with logistic regression correctly ranking fire risk in 99.8% of cases. --- ## Features Used | Feature | Description | |---|---| | `Temp` | Temperature (°C) | | `RH` | Relative humidity (%) | | `Ws` | Wind speed (km/h) | | `Rain` | Rainfall (mm) | | `FFMC` | Fine Fuel Moisture Code — ignition potential of small surface fuels | | `DMC` | Duff Moisture Code — moisture in deeper organic material | | `DC` | Drought Code — long-term drying of deep organic layers | | `ISI` | Initial Spread Index — predicted rate of fire spread | | `BUI` | Build Up Index — measure of total fuel available | | `FWI` | Fire Weather Index — overall fire intensity rating | | `Target` | Binary: `fire` or `notfire` | --- ## Key Findings - **Temperature and FFMC** are strongly correlated (r = 0.68) — hotter weather dries fine fuels, increasing ignition risk - **Humidity negatively correlates** with FFMC (r = -0.65), ISI (r = -0.69), and FWI (r = -0.58) — moisture suppresses flammability - **ISI is the strongest predictor of FWI** (r = 0.92), confirming its central role in fire intensity calculation - Fire days consistently show **higher median ISI and temperatures above 35°C**, a known critical threshold for Algerian wildfires --- ## P …