davis-mironga/Algerian_Forest_Fires_Classification

# Algerian Forest Fires Classification > Predicting forest fire occurrences in Algeria using machine learning and Fire Weather Index (FWI) components — a reproducible, end-to-end classification study. --- ## Table of Contents - Overview - Problem Statement - Project Structure - Dataset - Methodology - Key Findings - Quick Start - Experimental Results - Visuals - Future Work - Citation - Author --- ## Overview Forest fires pose a significant environmental and humanitarian risk, particularly across the Mediterranean and North African regions. Early, accurate prediction of fire risk can support timely intervention and reduce ecological damage. This project builds and evaluates binary classification models to predict whether a forest fire will occur on a given day in one of two Algerian regions, based on daily meteorological observations and Canadian Forest Fire Weather Index (FWI) system components. **Target Audience:** Data scientists, environmental researchers, machine learning practitioners, and academic researchers interested in fire risk modelling and reproducible ML workflows. --- ## Problem Statement Predicting forest fires requires models that can handle multi-source weather features and generalise across geographically distinct regions. While the FWI system provides established fire danger indices, it does not directly output a binary fire/no-fire classification. This project addresses the following question: > Can machine learning classifiers, trained on daily weather and FWI features, accurately distinguish fire-prone days from non-fire days across two climatically distinct Algerian regions? --- ## Project Structure ``` Algerian_Forest_Fires_Classification/ │ ├── data/ │ ├── raw/ │ │ └── Algerian_forest_fires_dataset_UPDATE.xlsx # Original UCI dataset │ └── processed/ │ └── cleaned_algerian_forest_fires.csv # Cleaned & preprocessed data │ ├── models/ │ ├── log_reg_model.pkl # Saved Logistic Regression …