This project predicts the Fire Weather Index (FWI) for the Algerian Forest Fire Dataset using machine learning.
It includes a Flask web application that allows users to input weather parameters (Temperature, RH, Wind Speed, Rain, etc.) and get a predicted FWI instantly.
Forest fires are among the most devastating natural disasters, and early prediction can help prevent massive environmental losses.
This project focuses on predicting Fire Weather Index (FWI) using meteorological parameters from the Algerian Forest Fire dataset.
After testing multiple regression algorithms, Ridge Regression was chosen because it provided the best performance and least overfitting compared to other models.
- End-to-end Machine Learning pipeline
- Data cleaning and EDA using Jupyter notebooks
- Model training with Ridge Regression
- Scaled data using StandardScaler
- Flask-based web app for user input and prediction
- Interactive form for entering temperature, humidity, wind speed, rainfall, etc.
- Instant FWI prediction on the same page (no reload)
| Category | Tools / Libraries |
|---|---|
| Programming | Python |
| Data Handling | Pandas, NumPy |
| Model | Ridge Regression (scikit-learn) |
| Visualization | Matplotlib, Seaborn |
| Web Framework | Flask |
| Frontend | HTML, CSS, JavaScript |
| Deployment | Localhost / GitHub integration ready |
Algerian Forest Fire Dataset
Contains meteorological variables:
- Temperature (°C)
- Relative Humidity (%)
- Wind Speed (Km/h)
- Rain (mm)
- FFMC, DMC, DC, ISI, BUI
- Region (Bejaia, Sidi Bel-abbes)
- Class (Fire / Not Fire)
Dataset Source: UCI Machine Learning Repository
| Category | MAE | R2 Score |
|---|---|---|
| Linear Regression | 0.54 | 0.98 |
| Lasso Regression | 1.13 | 0.94 |
| Ridge Regression | 0.56 | 0.984 |
✅ Ridge Regression was selected because it generalized best on unseen data.
Harsh Patil 📧 [harshrp2309@gmail.com]