This project involves building a classification model to predict the occurrence of forest fires (Fire or Not Fire) using the Algerian Forest Fires Dataset. It includes data cleaning, EDA, feature engineering, and model development. The project was implemented as part of a Udemy course by Krish Naik.
The dataset includes 244 instances of forest weather data from two regions in Algeria:
- Bejaia Region (northeast Algeria)
- Sidi Bel-abbes Region (northwest Algeria)
Each region contributes 122 instances, covering the period from June to September 2012. The dataset is suitable for both regression and classification tasks. This project focuses on the classification task — predicting fire occurrence.
| # | Feature Name | Description | Range |
|---|---|---|---|
| 1 | Date | Day, Month, Year (DD/MM/YYYY) | June–September 2012 |
| 2 | Temp | Noon Temperature (°C) | 22 – 42 |
| 3 | RH | Relative Humidity (%) | 21 – 90 |
| 4 | Ws | Wind Speed (km/h) | 6 – 29 |
| 5 | Rain | Total Daily Rain (mm) | 0 – 16.8 |
| 6 | FFMC | Fine Fuel Moisture Code | 28.6 – 92.5 |
| 7 | DMC | Duff Moisture Code | 1.1 – 65.9 |
| 8 | DC | Drought Code | 7 – 220.4 |
| 9 | ISI | Initial Spread Index | 0 – 18.5 |
| 10 | BUI | Buildup Index | 1.1 – 68 |
| 11 | FWI | Fire Weather Index | 0 – 31.1 |
| 12 | Classes | Fire occurrence (Target) | 1 = Fire, 0 = Not Fire |
- EDA: Visualized distributions, trends, and relationships among features.
- Data Cleaning: Handled missing values, standardized formats, and encoded categorical features.
- Feature Engineering: Created new inputs and normalized features for model training.
- Model Training: Trained a classification model (e.g., Gradient Boosting) to predict fire occurrence.
- Deployment: Built a simple Flask web app for fire occurrence prediction based on user inputs.
This project was built while following the Machine Learning course by Krish Naik on Udemy. It helped me gain hands-on experience with classification problems, web deployment, and end-to-end ML workflows.