TY - JOUR

T1 - Crosstown traffic

T2 - supervised prediction of impact of planned special events on urban traffic

AU - Tempelmeier, Nicolas

AU - Dietze, Stefan

AU - Demidova, Elena

N1 - Funding Information: This work was partially funded by the Federal Ministry of Education and Research (BMBF) under the project ?Data4UrbanMobility?, grant ID 02K15A040.

PY - 2020/4

Y1 - 2020/4

N2 - Large-scale planned special events in cities including concerts, football games and fairs can significantly impact urban mobility. The lack of reliable models for understanding and predicting mobility needs during urban events causes issues for mobility service users, providers as well as urban planners. In this article, we tackle the problem of building reliable supervised models for predicting the spatial and temporal impact of planned special events with respect to road traffic. We adopt a supervised machine learning approach to predict event impact from historical data and analyse effectiveness of a variety of features, covering, for instance, features of the events as well as mobility- and infrastructure-related features. Our evaluation results on real-world event data containing events from several venues in the Hannover region in Germany demonstrate that the proposed combinations of event-, mobility- and infrastructure-related features show the best performance and are able to accurately predict spatial and temporal impact on road traffic in the event context in this region. In particular, a comparison with both event-based and event-agnostic baselines shows superior capacity of our models to predict impact of planned special events on urban traffic.

AB - Large-scale planned special events in cities including concerts, football games and fairs can significantly impact urban mobility. The lack of reliable models for understanding and predicting mobility needs during urban events causes issues for mobility service users, providers as well as urban planners. In this article, we tackle the problem of building reliable supervised models for predicting the spatial and temporal impact of planned special events with respect to road traffic. We adopt a supervised machine learning approach to predict event impact from historical data and analyse effectiveness of a variety of features, covering, for instance, features of the events as well as mobility- and infrastructure-related features. Our evaluation results on real-world event data containing events from several venues in the Hannover region in Germany demonstrate that the proposed combinations of event-, mobility- and infrastructure-related features show the best performance and are able to accurately predict spatial and temporal impact on road traffic in the event context in this region. In particular, a comparison with both event-based and event-agnostic baselines shows superior capacity of our models to predict impact of planned special events on urban traffic.

KW - Event impact

KW - Planned special events

KW - Road traffic

KW - Urban mobility

UR - http://www.scopus.com/inward/record.url?scp=85066139081&partnerID=8YFLogxK

U2 - 10.1007/s10707-019-00366-x

DO - 10.1007/s10707-019-00366-x

M3 - Article

AN - SCOPUS:85066139081

VL - 24

SP - 339

EP - 370

JO - GEOINFORMATICA

JF - GEOINFORMATICA

SN - 1384-6175

IS - 2

ER -