---
license: mit
tags:
- disaster-prediction
- risk-assessment
- tabular-regression
datasets:
- emirhanakku/disaster-events-2025
---

# Disaster Risk Prediction Model

This model predicts disaster risk scores (0-1) based on location and disaster type.

## Model Details

- **Model Type:** Random Forest Regressor
- **Framework:** scikit-learn
- **Dataset:** Disaster Events 2025 (Kaggle)
- **Features:** Location (encoded), Disaster Type (encoded)
- **Target:** Risk Score (0.0 to 1.0)

## Usage

```python
import joblib
from huggingface_hub import hf_hub_download

# Download model
model_path = hf_hub_download(
    repo_id="IVB-2005/disaster-model",
    filename="disaster_risk_model.pkl"
)
model = joblib.load(model_path)

# Download encoders
disaster_enc_path = hf_hub_download(
    repo_id="IVB-2005/disaster-model",
    filename="disaster_encoder.pkl"
)
disaster_encoder = joblib.load(disaster_enc_path)

location_enc_path = hf_hub_download(
    repo_id="IVB-2005/disaster-model",
    filename="location_encoder.pkl"
)
location_encoder = joblib.load(location_enc_path)

# Make prediction
location_encoded = location_encoder.transform(['India'])[0]
disaster_encoded = disaster_encoder.transform(['Earthquake'])[0]

features = [[location_encoded, disaster_encoded]]
risk_score = model.predict(features)[0]

print(f"Risk Score: {risk_score:.3f}")
```

## Risk Levels

- **0.0 - 0.3:** LOW risk
- **0.3 - 0.7:** MEDIUM risk
- **0.7 - 1.0:** HIGH risk

## Training Data

Trained on real disaster events from 2025 including:
- Earthquakes
- Hurricanes
- Volcanic Eruptions
- Landslides
- Wildfires
- Droughts

## Performance

- Mean Absolute Error (MAE): ~0.05
- R² Score: ~0.85

## License

MIT License - Free for educational and commercial use.

## Citation

```bibtex
@misc{disaster-risk-model,
  author = {Your Name},
  title = {Disaster Risk Prediction Model},
  year = {2025},
  publisher = {Hugging Face},
  url = {https://huggingface.co/IVB-2005/disaster-model}
}
```