This end-to-end machine learning project focuses on predicting the Fire Weather Index (FWI) using the Algerian Forest Fires dataset. The goal is to help anticipate wildfire risk based on weather and environmental patterns from two regions in Algeria.
The dataset contains daily meteorological data from two regions:
- Bejaia Region
- Sidi Bel-Abbes Region
Each entry includes features like temperature, wind speed, humidity, rain, and corresponding FWI values.
- Merged region-wise data into a single structured dataset
- Removed irrelevant columns and handled missing values
- Standardized column names and converted dates
- Encoded region and date-based features for modeling
- Correlation heatmaps to identify linear relationships
- Pair plots and KDE plots for feature distribution analysis
- Line plots for temperature, wind speed, and humidity trends over time
- Regional comparison of weather patterns and fire risk
- Detected multicollinearity using heatmaps
- Created additional features to improve prediction accuracy
- Trained and evaluated the following regression models:
- Linear Regression
- Ridge, Lasso , ElasticNet Regression
- Model Pickling
- Evaluated models using:
- R² Score
- Mean Absolute Error (MAE)
- Root Mean Squared Error (RMSE)
- Correlation matrix and heatmaps
- Region-wise line plots for environmental variables
- Actual vs. predicted FWI plots
- Feature importance charts from ensemble models
- Python
- Pandas, NumPy
- Matplotlib, Seaborn
- Scikit-learn
- Jupyter Notebook
- Deploying the model with Flask or Streamlit for public use
- Implementing time-series forecasting for FWI trends
- Adding real-time data ingestion for dynamic prediction
Click here to view the project
Yukti Garg
Passionate about machine learning, data science, and solving real-world problems through data.