GeoTwin-RAG: A Digital Twin Framework for Robust Traffic Forecasting and Risk-Aware Tourism Mobility in Aseer Region
Mathematics · 14.11.2026 · 14 (22) · 4187
DOI: 10.3390/math14224187 ↗Kurz gefasst
Fachartikel in der Zeitschrift Mathematics, online erschienen am 14.11.2026. Verfasst von Abdulaziz Almaleh und 1 weiteren Autor:innen. Die Arbeit nennt 52 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: King Khalid University.
Themen: Mobilisation
Abstract (Originalsprache)
Mountain-tourism mobility is influenced by interacting factors including terrain, weather, visitor demand, network disruption, short-horizon traffic dynamics, and uncertainty in route decisions. This study presents GeoTwin-RAG, an auditable mobility decision framework that integrates an open-data geospatial digital twin of Aseer Region with microscopic traffic simulation, leakage-safe spatiotemporal forecasting, uncertainty- and risk-aware routing, and post-decision evidence retrieval. The experimental design comprises five scenarios, five demand multipliers, and five stochastic seeds, yielding 125 SUMO runs and 1,725,079 observed dynamic edge–time records across 1392 corridor edges. On a common four-window test subset of 182,283 observations, GraphSAGE–GRU achieved the best point-estimate performance with a mean absolute error (MAE) of 0.854 s, RMSE of 2.493 s, and R2=0.9893, although confidence intervals overlapped those of other leading models. Demand out-of-distribution testing increased MAE to 0.911 s, whereas scenario holdout increased it to 1.298 s, indicating greater sensitivity to structural operating-regime changes than to demand scaling alone. Routing was evaluated over 81 uncertainty–risk weight combinations and 497 feasible origin–destination–scenario cases, producing route-change rates from 0.20% to 74.85%. System ablation showed that modeled dynamic risk was the dominant driver of route adaptation: The travel-time-plus-risk configuration changed 60.76% of routes while reducing mean modeled risk by approximately 4.80%. Route recommendations were highly stable across stochastic seeds, with a mean edge-set Jaccard similarity of 0.994. The route-conditioned evidence layer achieved 100% paragraph-level citation coverage with no unknown evidence identifiers. External plausibility assessment used 13 raw HERE traffic snapshots collected across seven local dates, corresponding to 11 temporal clusters after consolidating near-simultaneous observations. These data are used for calibration and sensitivity analysis rather than as comprehensive field validation. Overall, GeoTwin-RAG provides a reproducible framework linking simulation, forecasting, explicit routing trade-offs, robustness analysis, external traffic reference data, and traceable decision explanation.
Förderung
- King Khalid University
Zitieren
APA
Almaleh, A., & Alqahtani, A. M. (2026). GeoTwin-RAG: A Digital Twin Framework for Robust Traffic Forecasting and Risk-Aware Tourism Mobility in Aseer Region. Mathematics, 14(22), 4187. https://doi.org/10.3390/math14224187
Vancouver
Almaleh A, Alqahtani AM. GeoTwin-RAG: A Digital Twin Framework for Robust Traffic Forecasting and Risk-Aware Tourism Mobility in Aseer Region. Mathematics. 2026;14(22):4187. doi:10.3390/math14224187
BibTeX
@article{Almaleh2026,
author = {Almaleh, Abdulaziz and Alqahtani, Abdullah M.},
title = {GeoTwin-RAG: A Digital Twin Framework for Robust Traffic Forecasting and Risk-Aware Tourism Mobility in Aseer Region},
journal = {Mathematics},
year = {2026},
volume = {14},
number = {22},
pages = {4187},
doi = {10.3390/math14224187}
}
EUVIMED