Machine Learning / Project
European energy demand forecasting
Electricity demand forecasting with XGBoost and LightGBM, connected to an API and model lifecycle tooling.
- My role
- Machine learning pipeline development
- Period
- Nov 2025 — Jan 2026
- R²
- 0.85+
- MAPE
- <10%
Project evaluation figures; performance depends on the time period, forecast horizon and split.
01 / The context
What the project addresses
A forecasting pipeline for European electricity demand using the Western Europe Power Consumption dataset.
02 / My contribution
What I worked on
Built the forecasting workflow and connected model training, prediction serving, time-series storage and retraining.
The challenge
Bring forecasting models into a repeatable data-to-prediction workflow.
03 / The approach
How the pieces fit together
- Prepare electricity demand time series for forecasting.
- Train and evaluate XGBoost and LightGBM models.
- Expose predictions through FastAPI and version models with MLflow.
- Use TimescaleDB for time-series storage and Airflow for ingestion and retraining workflows.
04 / The outcome
What the work delivered
- An end-to-end forecasting pipeline connecting data ingestion, model evaluation and API predictions.
Next case study