Citation

Pavlyuk D, Spiridovska N, Yatskiv (Jackiva) I. Transport 2020; 35(6): 576-587.

Copyright

(Copyright © 2020, Vilnius Gediminas Technical University and Lithuanian Academy of Sciences, Publisher Vilnius Gediminas Technical University (VGTU) Press)

DOI

10.3846/transport.2020.14159

PMID

unavailable

Abstract

Sustainable urban mobility remains an emerging research topic during last decades. In recent years, the smart card data collection systems have become widespread and many studies have been focused on usage of anonymized data from these systems for better understanding of mobility patterns of Public Transport (PT) passengers. Data-driven mobility patterns can benefit transport planners at strategic, tactical, and operational levels. A particular point of interest is a spatiotemporal dynamics of mobility patterns that highlights transformation of the PT passenger flows over the time continuously or in response to modifications of the PT system and policies. This study is aimed to estimation and analysis of the spatiotemporal dynamics of PT passenger flows in Riga (Latvia). A multi-stage methodology was proposed and includes three main stages: (1) estimation of individual trip vectors, (2) clustering of trip vectors into spatiotemporal mobility patterns, and (3) further analysis of mobility patterns' dynamics. The best practice methods are applied at every stage of the proposed methodology: the smart card validation flow is used for extracting information on boarding locations; the trip chain approach is used for estimation of individual trip destinations; vector-based clustering algorithms are utilised for identification of mobility patterns and discovering their dynamics. The resulting methodology provides an advanced tool for observing and managing of PT demand fluctuation on a daily basis. The methodology was applied for mining of a large smart card data set (124 million records) for year 2018. Most important empirical results include obtained daily mobility patterns in Riga, their clusters, and within-cluster dynamics over the year. Obtained daily mobility patterns allows estimation of a city-level PT origin-destination matrix that is useful in many applied areas, e.g., dynamic passenger flow assignment models. Mobility pattern-based clustering of days allows effective comparison and flexible tuning of the PT system for different days of a week, public holidays, extreme weather conditions, and large events. Dynamics of mobility patterns allows estimating the effect of implementing changes (e.g., fare increase or road maintenance) and demand forecasting for user-focused development of PT system.
First published online 6 January 2021

Language: en

Keywords

big data; clustering; public transport; smart card data; transport modelling; user travel behaviour