saichaitany0405/Forest_Fire_Prediction

End-to-end Machine Learning project for **Forest Weather Index (FWI) prediction** using the Algerian Forest Fires dataset. Includes data cleaning, feature engineering, Linear Regression modeling with StandardScaler, model evaluation (MSE, MAE, RMSE, R²), serialization using pickle, and deployment via a Flask web app for real-time FWI prediction. # Forest Weather Index (FWI) Prediction This project is an end-to-end Machine Learning application that predicts the **Forest Weather Index (FWI)** using meteorological and fire-related inputs. It covers data preprocessing, feature engineering, model training, evaluation, serialization, and deployment using Flask. --- ## Project Description The Forest Weather Index (FWI) is a critical component of the Canadian Forest Fire Danger Rating System and is used to estimate the intensity of forest fires. This application allows users to input weather conditions and fire indicators through a web interface and obtain a real-time FWI prediction. --- ## Dataset **Dataset Used:** Algerian Forest Fires Dataset **Source File:** `datasets/Algerian_forest_fires_dataset_UPDATE.csv` ### Data Preprocessing - Removed null values and invalid rows - Stripped and cleaned column names - Converted features to appropriate numeric data types - Encoded categorical variables: - `Classes`: not fire → 0, fire → 1 - `region`: Bejaia → 0, Sidi-Bel Abbes → 1 - Removed date columns (`day`, `month`, `year`) - Saved cleaned data for reuse **Cleaned Dataset:** `datasets/cleaned_dataset.csv` --- ## Features and Target ### Input Features - Temperature - RH (Relative Humidity) - Ws (Wind Speed) - Rain - FFMC - DMC - DC - ISI - BUI - Classes - region ### Target Variable - FWI (Forest Weather Index) --- ## Model Details - **Algorithm:** Linear Regression - **Feature Scaling:** StandardScaler - **Train-Test Split:** 80% training / 20% testing - **Serialization:** Model and scaler saved using `pickle` --- ## Model Performance | Metric | Value | |--------|-------| | Mean Squared Error (MSE) | 0.3323 | | Mean Absolute Error (MAE) | 0.4268 | | Root Mean Squared Error (RMSE) | 0.5765 | | R² Score | 0.9890 | An R² score of **0.989** indicates that the model explains approximately **98.9% of the variance** in the Forest Weather Index. --- ## Project Structure ```text Forest_fire_prediction/ ├── …