2025-12-09 –, Auditorium
Have you ever happened to use GPS and realised that it is not working properly? The Sun and a Space Weather effect called Travelling Ionospheric Disturbances (TIDs) could be responsible. We will present an explainable TIDs forecasting model, based on CatBoost and using several physical drivers to make forecasts.
Travelling Ionospheric Disturbances (TIDs) are wave-like fluctuations in plasma density that ripple through the ionosphere, excited mainly by acoustic and gravity waves. TIDs contribute significantly to the degradation of Global Navigation Satellite Systems (GNSS) and radio communications at mid-latitudes, so it is crucial to reliably detect and potentially forecast their occurrence.
Since there are no empirical models to forecast TIDs, in the context of the Travelling Ionospheric Disturbances Forecasting System (T-FORS) project, funded by the European Union, Istituto Nazionale di Geofisica e Vulcanologia (INGV) developed an explainable machine learning model, which is based on gradient boosting over an ensemble of decision trees (CatBoost) and aims at forecasting the occurrence of TIDs over the European sector within 3 hours.
The SHapley Additive exPlanation (SHAP) framework is introduced to test features influence on the model output from both global and local (event-level) aspects, so as to improve interpretability and explainability, which are very desirable features in potentially high-risk settings such as GNSS and HF communications.
Our research shows that the proposed algorithm, in its current development stage, achieves a reasonable level of performance – given the imbalanced nature of the problem and the still incomplete understanding of the physical phenomenon.
A results-driven data professional – focused on hype-free solutions tailored to business needs.
I am currently creating value at the National Institute of Geophysics and Volcanology (INGV), where I develop machine learning models in the Space Weather domain. My job is complemented by finding the hidden stories in data and make them accessible to stakeholders. I studied Physics in Italy (Napoli) and Germany (Frankfurt am Main), previously worked on Analytics in the strategic division of the world's largest professional services network, and in the Data Science department of the leading Italian publisher.
When not at work, I enjoy theatre, talking about finance or learning a new language.