Illia Babounikau
Dr. Illia Babounikau is an accomplished data scientist with extensive expertise in machine learning and forecasting. He holds a Ph.D. in Physics from Hamburg University and initially pursued an academic career, focusing on large-scale data analysis and machine learning applications. His contributions have been instrumental in international scientific collaborations, including the CMS experiment at CERN’s Large Hadron Collider and the COMET project at J-PARC.
For the past five years, Dr. Babounikau has been a Data Scientist at Blue Yonder and VOIDS, specializing in developing and fine-tuning advanced forecasting models for retail planning and inventory management. He leads the design and implementation of tailored machine-learning solutions, addressing complex challenges within supply chains across diverse industries.
Dr. Babounikau is passionate about bridging the gap between data science and business strategy, ensuring machine learning models are aligned with business objectives to drive data-informed decision-making.
Session
Forecasting talks love a clean ending: “and then we improved WMAPE by 3.7%.”
Nice. Now put that model into production without suffering from instability.
You retrain your model on a few new weeks of data and suddenly the one-year forecast jumps 15–20%. Planning teams redo decisions, trust erodes, and your “accurate” model becomes unusable. This talk is about forecast stability: how much forecasts change when you add new data and rerun the same pipeline.
We run a simple experiment: train a model, forecast one year ahead, add recent data, retrain, and measure forecast-to-forecast change. We repeat this across common forecasting approaches including ETS/ARIMA, Prophet, XGBoost with lag features, AutoGluon ensembles, neural/global models, and TimeGPT-style APIs.
You will see that high accuracy does not guarantee usable forecasts, and that some models are systematically more volatile than others. We then cover practical ways to stabilise forecasts without freezing them, focusing on reconciliation and ensembling (including origin ensembling).
This talk is for forecasting practitioners who want models users actually trust, not just good metrics.