BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//pretalx//programme.europython.eu//europython-2026//speaker//FRRAE
 7
BEGIN:VTIMEZONE
TZID:CET
BEGIN:STANDARD
DTSTART:20001029T040000
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=10
TZNAME:CET
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
END:STANDARD
BEGIN:DAYLIGHT
DTSTART:20000326T030000
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=3
TZNAME:CEST
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
END:DAYLIGHT
END:VTIMEZONE
BEGIN:VEVENT
UID:pretalx-europython-2026-VXDYGX@programme.europython.eu
DTSTART;TZID=CET:20260715T114000
DTEND;TZID=CET:20260715T121000
DESCRIPTION:**Space Weather** doesn’t just produce beautiful auroras: it 
 can silently disrupt navigation systems\, radio links\, and satellite-base
 d technologies we rely on every day.\n\nTravelling Ionospheric Disturbance
 s (TIDs) are wave-like structures in the ionosphere that affect GNSS accur
 acy and HF communications. From an ML perspective\, forecasting TIDs is a 
 challenging rare-event prediction problem involving imbalanced data and he
 terogeneous physical inputs.\n\nIn this talk\, I will present an operation
 al machine learning approach developed within the T-FORS project to foreca
 st TID occurrence over Europe. The model is built using **CatBoost** and i
 ntegrates data from space- and ground-based observations.\n\nThe talk focu
 ses on **model design and evaluation choices**. In particular\, I will sho
 w how **SHAP** can be used to debug model behaviour\, validate feature rel
 evance\, and build trust in predictions in a high-risk operational context
 .\n\nAlong the way\, I’ll share practical engineering lessons on:\n- han
 dling class imbalance\,\n- incorporating domain knowledge into ML pipeline
 s\,\n- producing **uncertainty-aware outputs** via **Conformal Prediction*
 *\, and\n- running **interpretable models in real-time forecasting systems
 **.\n\nThe talk is aimed at data scientists and ML practitioners intereste
 d in applied forecasting\, interpretable models\, uncertainty quantificati
 on and ML at the boundary between data and physics.\n\n---\n\n**Talk outli
 ne**\n- 0-4: What is Space Weather and why should we care\n- 4-7: Framing 
 TID forecasting as an ML problem\n- 7-10: Model design with CatBoost\n- 10
 -13: Explainability with SHAP\n- 13-18: Uncertainty quantification with Co
 nformal Prediction\n- 18-22: Cost-sensitive learning and real-time operati
 ons\n- 22-25: Lessons learned\n- 25-30: Q&A
DTSTAMP:20260524T121708Z
LOCATION:Chamber Hall B (S3B)
SUMMARY:When the Sun Breaks Your GPS: Building an Explainable Early Warning
  System - Vincenzo Ventriglia
URL:https://programme.europython.eu/europython-2026/talk/VXDYGX/
END:VEVENT
END:VCALENDAR