At the SEM4TRA 2025 workshop of the SEMANTICS conference, Adrian M.P. Brasoveanu (MODUL University Vienna / MODUL Technology) presented AI-CENTIVE’s latest research on Explainable Mobility Prediction in Urban Transit Zones.
The talk highlighted how hybrid AI models – combining ensemble neural networks, statistical methods, and Graph Neural Networks (GNNs) – can predict travel behavior more accurately by modeling relationships between trips, routes, and activity types.
A key achievement: the system not only delivers personalized mobility recommendations, but also provides transparent explanations for why a route was suggested. This is enabled by a semantic pipeline that translates model outputs into human-readable justifications. This semantic layer also accelerates large-scale analysis, making queries up to five times faster than with conventional data science tools.
Results from the second user pilot (March–May 2025) showed that participants responded positively to notifications when these were backed by confidence scores and contextual reasoning (e.g., familiarity of locations, time patterns, activity types).
The team is now preparing a journal article with the full results and plans to reuse the semantic pipeline in future projects where explainability at scale is crucial.
Ontologies are also available online!
