Transport Policy 14 (2007) 193–203 Measuring transit use variability with smart-card data Catherine Morency b,c,Ã , Martin Tre´panier a,b,c , Bruno Agard a,c a Groupe Polygistique, E ´ cole Polytechnique de Montre´al, C.P. 6079, succ. Centre-ville, Montre´al, Que., Canada H3C3A7 b Groupe MADITUC, E ´ cole Polytechnique de Montre´al, C.P. 6079, succ. Centre-ville, Montre´al, Que., Canada H3C3A7 c Centre Interuniversitaire de Recherche sur les Re´seaux d’Entreprise, la Logistique et le Transport (CIRRELT), E ´ cole Polytechnique de Montre´al, C.P. 6079, succ. Centre-ville, Montre´al, Que., Canada H3C3A7 Available online 12 March 2007 Abstract The potential of smart-card data for measuring the variability of urban public transit network use is the focus of this paper. Data collected during 277 consecutive days of travel on a Canadian transit network are processed for this purpose. The organization of data using an object-oriented approach is discussed. Then, measures of spatial and temporal variability of transit use for various types of card are defined and estimated using the data sets presented. Data mining techniques are also used to identify transit use cycles and homogenous days and weeks of travel among card segments and at various times of the year. r 2007 Elsevier Ltd. All rights reserved. Keywords: Smart cards; Transit system; Travel behaviour; Data mining; Variability 1. Introduction In most urban networks, the demand for public transit constantly changes, depending on the time of travel (day of the week, season or holiday) and other factors like weather and service breakdown. Often, transit operators find it extremely difficult to adjust the service to the demand, and, clearly, better adjustment could reduce operating costs and help optimize vehicle use over the network. One of the main issues is the ability to measure the demand precisely and understand its dynamics in order to establish day-to- day predictions. Today, tools are available to planners to perform this task. The purpose of this study is to illustrate the potential of smart-card data to measure the spatial and temporal variability of transit use. In order to do this, the object- oriented approach, data mining techniques and database management tools are used to construct systematic indicators that help evaluate the variability of travel behaviours by various population segments on a transit network. The paper is organized as follows. First, a review of the literature in the relevant research fields is provided. Smart- card data systems and the processing of outputted data sets are discussed, as well as the potential of data mining techniques for various applications. The evaluation and measurement of the variability of travel behaviour are also discussed. The data set available for the analysis is then described, and technical details regarding its collection and processing are provided. The measurement concepts and methods are then presented. These relate to the spatial and temporal indices defined to measure the variability of travel behaviour on a transit network. The results of the analysis are subsequently presented, followed by a discus- sion and some future research avenues drawing on insights gained from the current research. 2. Literature review 2.1. Smart-card data The use of smart-card automated fare collection systems in public transit is spreading throughout the world. Even ARTICLE IN PRESS www.elsevier.com/locate/tranpol 0967-070X/$ - see front matter r 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.tranpol.2007.01.001 Ã Corresponding author. E ´ cole Polytechnique de Montre´al, C.P. 6079, succ. Centre-ville, Montre´al, Que., Canada H3C3A7. Tel.: +1 514 340 4711x4502; fax: +1 514 340 5763. E-mail addresses: cmorency@polymtl.ca (C. Morency), mtrepanier@polymtl.ca (M. Tre´panier), bruno.agard@polymtl.ca (B. Agard).