A model for measuring activity similarity between public transit passengers using smart card data. (October 2018)

• Record Type: Journal Article
• Authors: Faroqi, Hamed; Mesbah, Mahmoud; Kim, Jiwon; Tavassoli, Ahmad
• Is Part Of: Travel behaviour and society. Volume 13(2018)
• Page Start: 11
• Page End: 25
• Publication Date: 2018-10
• Subjects: Time-geography -- Data mining -- Travel behaviour -- Planning -- Spatiotemporal; Transportation -- Periodicals; Population geography -- Periodicals; 303.48305
• DOI: 10.1016/j.tbs.2018.05.004
• Languages: English
• ISSNs: 2214-367X

Abstract

Highlights: The STP method is deployed to the transit smart card data. Time is considered as a continuous attribute of the activity. Three main elements of the activity are considered in the model. The model focuses on the passengers. Passengers behavior during their activities are focused. Abstract: An activity is characterized by its location, time and type. Smart card data include the location and time of boarding and/or alighting transactions within the public transit system. This data can be used to study the spatiotemporal range of the activity as it usually happens between an alighting and the next boarding transaction. This kind of activity can also be inferred from the start time and duration of the activity, and the available land use in the vicinity. This paper proposes a model which considers the three main characteristics of the activity to measure similarities between passengers' activities. The model consists of two parallel steps—one for the spatiotemporal aspects and the other for the activity type. The first one uses the concept of Space Time Prism (STP) to measure the spatiotemporal similarity of two activities in a three-dimensional continuous space. The latter models the activity type using a probabilistic decision tree to measure the activity type similarity. The final activity similarity value is the product of the activity type and the spatiotemporal similarity values. The model is implemented for four-day smart card data in Brisbane, Queensland [Australia]. In order to confirm the results of the model, the passengers are clustered and discussed based on the measured activity similarity. The results show more than 81 per cent of the passengers have partial or complete activity similarity with their fellow passengers.