Susovan88/Algerian_forest_fires_app

# Algerian Forest Fires FWI Prediction App This web application predicts the Forest Fire Weather Index (FWI) for Algerian forests using environmental parameters. It is built with Flask and uses a machine learning pipeline (feature scaling, feature selection, and Lasso regression) to provide accurate FWI predictions. --- ## Table of Contents - Features - How It Works - How to Build the Model - Setup Instructions - File Structure - Environmental Parameters - Intended Users - License --- ## Features - **User-friendly web interface** for entering environmental data. - **Machine Learning Model**: Linear Regression with Exhaustive Feature Selection (EFS) and Lasso Regression. - **Responsive and mobile-friendly design**. - **Instant FWI prediction** for fire risk assessment and early warning. - **Educational and research tool** for environmental scientists and students. --- ## How It Works 1. **Input**: User enters environmental parameters (Temperature, Relative Humidity, Wind Speed, Rainfall, FFMC, DMC, ISI, Classes, Region). 2. **Processing**: Data is scaled and the best features are selected using EFS. 3. **Prediction**: The Lasso regression model predicts the FWI. 4. **Output**: The predicted FWI is displayed on the page. --- ## How to Build the Model To reproduce or improve the model, follow these steps: 1. **Data Collection** - Download the Algerian Forest Fires dataset (e.g., from UCI Machine Learning Repository. - Combine and clean the data as needed. 2. **Data Preprocessing** - Handle missing values and outliers. - Encode categorical variables (e.g., Classes, Region) if necessary. - Split the data into features (X) and target (FWI). 3. **Feature Scaling** - Use `StandardScaler` from scikit-learn to scale the features. - Save the scaler using `pickle` for later use in the app. 4. **Feature Selection** - Apply Exhaustive Feature Selection (EFS) to select the most relevant features. - You can use `mlxtend`'s `ExhaustiveFeatureSelector` or similar t …