atul-aiml/algerian_forest_fires_fwi_prediction
Domain:
environment and energyRecord type:
dataset
Creator:
atu
Host:
Predicting Fire Weather Index using Ridge/Lasso/ElasticNet regression
# Algerian Forest Fires – Fire Weather Index (FWI) Prediction
Regression project that predicts the **Fire Weather Index (FWI)** for forest fires in Algeria using weather and fire-behavior data, with a full pipeline from data cleaning to model selection.
## 📊 Dataset
The dataset regroups **244 instances** from two regions of Algeria:
- **Bejaia region** (northeast Algeria) — 122 instances
- **Sidi Bel-abbes region** (northwest Algeria) — 122 instances
Data covers **June to September 2012**. Each instance includes weather observations (temperature, relative humidity, wind speed, rain) and fire weather indices calculated from the Canadian FWI System (FFMC, DMC, DC, ISI, BUI, FWI), plus a fire/not-fire classification label.
**Files:**
- `Algerian_forest_fires_dataset_UPDATE.csv` — raw/original dataset
- `Algerian_forest_fires_cleaned_dataset.csv` — cleaned dataset used for modeling
## 🎯 Objective
Predict the **FWI (Fire Weather Index)** — a continuous score indicating fire intensity risk — using the other weather and fire-index features.
## 🛠️ Workflow
1. **Data Cleaning** (`Ridge__Lasso_Regression.ipynb`)
- Merged and cleaned two-region dataset
- Handled missing values, fixed column types
- Added a `Region` indicator column
- Encoded the `Classes` target (fire / not fire)
2. **Exploratory Data Analysis**
- Class distribution analysis
- Correlation heatmaps to detect multicollinearity
3. **Feature Engineering** (`Model_Training.ipynb`)
- Dropped date columns (day/month/year)
- Removed highly correlated features (threshold > 0.85)
- Standardized features using `StandardScaler`
4. **Modeling**
- Linear Regression (baseline)
- Lasso Regression + LassoCV (cross-validated)
- Ridge Regression + RidgeCV (cross-validated)
- ElasticNet Regression + ElasticNetCV (cross-validated)
- Evaluated using **MAE** and **R² Score**
## 📁 Repository Structure
```
├── Algerian_forest_fires_dataset_UPDATE.csv
├── Algerian_forest_fires_cleaned_dataset.csv
├── Ridge__Lasso_Regressi …