Proxpekt/Algerain_Forest_Fire_Predictor

The Algerian Forest Fire Predictor GitHub repo typically describes a machine learning project that uses historical weather data to predict the likelihood of a forest fire (classification) or the Fire Weather Index (FWI) (regression) in Algeria, often using models like Random Forest or Ridge Regression. # 🔥 Algerian Forest Fire Prediction using Machine Learning This project develops and evaluates several linear regression models to accurately predict the **Fire Weather Index (FWI)** — a critical measure of forest fire risk — using meteorological data from Algerian forest regions. The model aims to assist in **fire prevention strategies** and **efficient resource allocation**. --- ## 📋 Project Overview ### 🎯 Objective The primary goal is to **predict the continuous-variable Fire Weather Index (FWI)** using environmental and meteorological factors. This is formulated as a **Supervised Regression Problem**. --- ## 📊 Dataset **Dataset Used:** Algerian Forest Fires Dataset (UCI Repository) - **Regions Covered:** Bejaia (Northeast) and Sidi-Bel Abbes (Northwest), Algeria - **Time Period:** June – September 2012 - **Target Variable:** Fire Weather Index (**FWI**) ### 🌦️ Feature Overview | Feature Category | Example Attributes | |------------------|--------------------| | **Weather Data** | Temperature (Temp), Relative Humidity (RH), Wind Speed (Ws), Rain | | **FWI Components** | Fine Fuel Moisture Code (FFMC), Drought Code (DC), Initial Spread Index (ISI), Buildup Index (BUI) | --- ## 🛠️ Methodology ### 1. Data Cleaning & EDA (`data_cleaner.ipynb`) - **Initial Cleaning:** Fixed data entry errors, including a misplaced row combining two regions. - **Feature Engineering:** Converted categorical variable `Classes` ("fire", "not fire") into numeric labels (`1`, `0`) for analysis. - **Exploratory Data Analysis (EDA):** - Identified August–September as the **peak fire months** for both regions. - Visualized feature correlations to understand relationships between weather variables and fire risk. --- ### 2. Model Training & Selection (`model_training.ipynb`) - **Feature Selection:** - Removed `DC` due to high multicollinearity with `BUI` (correlation > 0.85). - **Data Preparation:** - Split dataset into **75% training** and **25% testing** sets. - Standardized fea …