Paolo Speziali, Mehrdad Asadi, Diederik M. Roijers, Ann Nowe
Personalized pedestrian routing can be framed as an instance of AI-driven adaptive routing, where the system learns user preferences and adjusts route recommendations accordingly. However, modeling heterogeneous route-choice preferences between users and travel contexts remains challenging, as fully individualized models are difficult to scale. We propose a context-aware routing framework that captures this heterogeneity through a small number of interpretable subgroup models. We apply Exceptional model mining to identify subgroups of observations exhibiting statistically distinct decision patterns relative to a global preference model. Our approach enables prediction through shared subgroup-specific preference models rather than individual optimization. Using simulated user models for context-dependent route choice decisions, our experimental results show that our approach successfully identifies meaningful context-dependent preference groups, supporting scalable and interpretable personalized pedestrian routing.
Speziali, P, Asadi, M, Roijers, DM & Nowe, A 2026, When Context Matters: Exceptional Model Mining for Pedestrian Route Choice. in Proceedings of the Fifth International Conference on Hybrid Human-Machine Intelligence. IOS Press.
Speziali, P., Asadi, M., Roijers, D. M., & Nowe, A. (Accepted/In press). When Context Matters: Exceptional Model Mining for Pedestrian Route Choice. In Proceedings of the Fifth International Conference on Hybrid Human-Machine Intelligence IOS Press.
@inproceedings{bcbf6621a46649249736e438deb52e14,
title = "When Context Matters: Exceptional Model Mining for Pedestrian Route Choice",
abstract = "Personalized pedestrian routing can be framed as an instance of AI-driven adaptive routing, where the system learns user preferences and adjusts route recommendations accordingly. However, modeling heterogeneous route-choice preferences between users and travel contexts remains challenging, as fully individualized models are difficult to scale. We propose a context-aware routing framework that captures this heterogeneity through a small number of interpretable subgroup models. We apply Exceptional model mining to identify subgroups of observations exhibiting statistically distinct decision patterns relative to a global preference model. Our approach enables prediction through shared subgroup-specific preference models rather than individual optimization. Using simulated user models for context-dependent route choice decisions, our experimental results show that our approach successfully identifies meaningful context-dependent preference groups, supporting scalable and interpretable personalized pedestrian routing.",
author = "Paolo Speziali and Mehrdad Asadi and Roijers, \{Diederik M.\} and Ann Nowe",
year = "2026",
month = may,
day = "1",
language = "English",
booktitle = "Proceedings of the Fifth International Conference on Hybrid Human-Machine Intelligence",
publisher = "IOS Press",
address = "Netherlands",
}