PetroPredict is an advanced analytics solution built on Microsoft Fabric that forecasts wholesale petrol prices in Poland. The system combines multiple data sources including oil prices, exchange rates, and historical petrol prices to generate accurate 12-month forecasts using SARIMAX (Seasonal AutoRegressive Integrated Moving Average with eXogenous variables) modeling.
- Multi-Source Data Integration: Automatically collects data from multiple APIs and web sources
- Advanced Time Series Forecasting: SARIMAX model with external factors for improved accuracy
- Real-time Data Pipeline: Automated data refresh and model retraining
- Interactive Visualizations: Power BI reports with historical trends and forecasts
- Confidence Intervals: 95% confidence bands for risk assessment
The solution is built on Microsoft Fabric and consists of:
-
Data Ingestion Notebooks
LoadFXRates: Fetches USD/PLN and EUR/PLN exchange rates from National Bank of Poland (NBP)LoadOilPrice: Retrieves Brent and WTI oil prices from Energy Information Administration (EIA)LoadPolishPetrolPrices: Scrapes historical petrol prices from Polish fuel retailers
-
Forecasting Engine
SARIMAX_Forecast_Lakehouse: Implements the forecasting model with feature engineering
-
Data Storage
- Microsoft Fabric Lakehouse with Delta tables for efficient querying
-
Visualization Layer
- Power BI semantic model and reports for business insights
| Source | Data Type | Update Frequency | History |
|---|---|---|---|
| NBP API | Exchange Rates (USD/PLN, EUR/PLN) | Daily | 1000 days |
| EIA API | Oil Prices (Brent, WTI) | Daily | Variable |
| Lotos Archives | Polish Petrol Prices | When available | Full history |
- Microsoft Fabric workspace
- Python environment with required packages
- API keys for:
- EIA (Energy Information Administration)
- Access to NBP API (no key required)
-
Environment Variables Create a
.envfile with:EIA_API_KEY=your_eia_api_key_here FABRIC_WORKSPACE_ID=your_workspace_id_here FABRIC_LAKEHOUSE_ID=your_lakehouse_id_here
-
Install Dependencies
pip install pandas numpy matplotlib seaborn requests beautifulsoup4 lxml pyspark
-
Configure Fabric Connection
- Import notebooks to your Fabric workspace
- Update lakehouse connections in each notebook
- Configure scheduled refresh for automated updates
-
Initial Data Load Execute notebooks in order:
1. LoadFXRates.Notebook 2. LoadOilPrice.Notebook 3. LoadPolishPetrolPrices.Notebook 4. SARIMAX_Forecast_Lakehouse.Notebook -
Scheduled Updates Configure Fabric pipeline to run daily for automated updates
- View forecasts in the Power BI report
- Query the
petrol_price_forecast_sarimaxtable for programmatic access - Export data to Excel/CSV for further analysis
- Seasonal Period: 7 days (weekly seasonality)
- External Regressors:
- Brent oil prices
- USD/PLN exchange rate
- EUR/PLN exchange rate
- Forecast Horizon: 365 days
- Confidence Level: 95%
The model is evaluated using:
- Mean Absolute Error (MAE)
- Root Mean Square Error (RMSE)
- Mean Absolute Percentage Error (MAPE)
We welcome contributions! Please:
- Fork the repository
- Create a feature branch (
git checkout -b feature/AmazingFeature) - Commit your changes (
git commit -m 'Add some AmazingFeature') - Push to the branch (
git push origin feature/AmazingFeature) - Open a Pull Request
- Follow PEP 8 for Python code
- Add unit tests for new features
- Update documentation for API changes
- Ensure all notebooks are properly documented
This project is licensed under the MIT License - see the LICENSE file for details.
- National Bank of Poland for exchange rate data
- U.S. Energy Information Administration for oil price data
- Lotos for historical petrol price data in Poland
- Microsoft Fabric team for the analytics platform
For questions or support, please open an issue in the repository.
Note: This project is for educational and analytical purposes. Always verify forecasts with domain experts before making business decisions.