Activity detection and transfer identification for public transit fare card data

This paper contributes to the emerging literature analysing public transit fare card data for a better understanding of passengers’ mobility patterns and path choices. A new heuristic is proposed to estimate the stop‐level origins and destinations by detecting the traveller activities in the observed transactions in a fare card dataset. The main focus in this research is estimating the actual passenger trajectories for multi‐leg journeys. If the fare card dataset includes both boarding and alighting information of each transaction, the main challenge is the estimation of origins and destinations by distinguishing the transfer interchanges from the activity locations. Built on commonly used criteria for identifying transfers, this paper proposes a new method to improve the accuracy of short activity detection to estimate the passengers’ true origins and destinations. The set of criteria in this research is based on the proposed concept of ‘off‐optimality’ for a more accurate identification of short/hidden activities within the labelled transfers. The measure of off‐optimality incorporates different variables of the transit service between the given journey ends (including alternative paths and routes, service headways, walk distances/times, transfer points, etc.) and reflects those into a simple quantity to improve the accuracy of estimation. In addition to off‐optimality, the time gap between two transactions, the total travel time, and the circuity of the path trajectories are other variables that are used in distinguishing the true transfers from activities. The proposed set of criteria is calibrated using a large endogenous set of fare card data from Brisbane, Australia, and is validated using a set of transit journeys and reported activities from a household travel survey. The results are presented for the fare card data from Brisbane collected in March 2013. The validation and case study results affirm the effectiveness of the proposed criteria in short activity detection.