umutssl/algerian-forest-fires--Ml

[EN] 🌲 Algerian Forest Fires — Machine Learning Project 🔥 Fire Weather Index (FWI) Prediction using Regression Models A machine learning approach to analyze and predict fire danger levels in Algeria’s forest regions. 📘 Project Overview This project aims to predict the Fire Weather Index (FWI) — a numerical rating representing fire danger — using various meteorological and environmental variables collected from Algerian forest regions. By leveraging regression algorithms and data preprocessing techniques, the project provides insight into how weather factors influence the risk of wildfires 🧩 Dataset Information Dataset Source: Algerian Forest Fires Dataset (UCI / Kaggle) Total Samples: 244 Regions: Bejaia Region: North-East Algeria (122 samples) Sidi Bel-Abbès Region: North-West Algeria (122 samples) Period: June 2012 – September 2012 Attributes: 11 input features + 1 output attribute (Classes) Target Variable: For regression: FWI (Fire Weather Index) For classification: Classes (fire / not fire) Class Distribution: 🔥 Fire: 138 instances 🌿 Not Fire: 106 instances The dataset includes meteorological variables such as temperature, relative humidity, wind speed, rainfall, and fire weather codes (FFMC, DMC, DC, ISI, BUI, FWI). These were collected daily across the two regions to analyze fire risk patterns. ⚙️ Data Preprocessing Data Cleaning Removed null and duplicate entries. Assigned region labels safely for Bejaia (0) and Sidi Bel-Abbes (1). Dropped irrelevant columns (e.g., date or text data). Feature Engineering Converted categorical values (Region) into numeric. Removed Classes column to prevent label leakage during FWI regression. Scaling Applied StandardScaler() to normalize numeric features. Train-Test Split 80% training, 20% testing. Ensured reproducibility with random_state=42. 🧠 Machine Learning Models Used | Model | Type | Cross V …