Assessing longitudinal stability of public transport users with smart card data

Many public transit networks around the world use the smart card data which provides the information about the users. In this regard, several methods are developed by mostly applying clustering approaches to perform data segmentation and discover the pattern of users. This study addresses the applicability of the temporal segmented data identified in 18 clusters for measuring the stability of users’ temporal habits as well as conducting descriptive analysis of the clusters, fare types and the days of the week to support the justification of findings. Each cluster contains users with their specific time and number of boardings. To understand whether the users are stable in the clusters, the sequential measurement based on the Euclidean distance between centres of the clusters, as the representatives of their members, is applied for each user over one month in this study. We ranked calculated measures to three different levels of high and medium stable or unstable using a histogram. The outcomes demonstrate the high stability of adult customers on three temporal routines, particularly regarding the days of the week. The users of the first and last working days of the week have a similar tendency in clusters’ membership tracks and pretty the same proportion of stability levels, having the minimum high stable and the maximum unstable users. Regarding the fare types, we recognized that regular students have the same unstable frequency in spite of having a significantly less frequency than regular.