subhadip-011/algerian-forest-fire-prediction

# Algerian Forest Fire Prediction Project This is my end-to-end Machine Learning project where I built a web application that can **predict the chances of a forest fire** in Algeria based on different weather and environmental features. I worked on data preprocessing, model training, evaluation, and deployment using Flask. ## Project Summary The main goal of this project is to help understand how different weather conditions (like temperature, humidity, wind, and rain) affect the possibility of forest fires. I used a public dataset from the UCI Machine Learning Repository which contains real meteorological data from two regions in Algeria — **Bejaia** and **Sidi Bel-abbes**. I trained and tested multiple regression models and finally selected the **Lasso Regression** model as it gave me the best results. ## Dataset Details **Dataset Name:** Algerian Forest Fires Dataset **Source:** UCI Machine Learning Repository **Features used in this project:** - Temperature (°C) - Relative Humidity (%) - Wind Speed (km/h) - Rain (mm) - Fine Fuel Moisture Code (FFMC) - Duff Moisture Code (DMC) - Drought Code (DC) - Initial Spread Index (ISI) - Class (Fire / Not Fire) The dataset has two regions of data — **Bejaia** and **Sidi Bel-abbes** — collected during the summer of 2012. ## Tools and Technologies - Python (v3.10) - Flask (for web app) - Pandas, NumPy (for data handling) - Scikit-learn (for ML algorithms) - Matplotlib, Seaborn (for visualization) - Pickle (for saving models) - HTML, CSS (for frontend) ## Model Comparison I trained three models and compared their R² scores: | Model | R² Score | |--------|-----------| | Linear Regression | 0.9497 | | Ridge Regression | 0.9501 | | **Lasso Regression** | **0.9543**| Based on this, I selected **Lasso Regression** as my final model since it performed the best. ## Project Workflow 1. **Data Preprocessing** - Cleaned missing values - Encoded labels and scaled numerical data using `StandardScaler` 2. **Model Training** …