Transfer journey identification and analyses from electronic fare collection data

Understanding the behaviour of public transport passengers is key to providing a system from which passengers will derive the maximum benefit. One method of analysing this behaviour is with the use of passenger boarding data, stored in a database. Such a database may be improved by enriching the already existing dataset by applying specific algorithms. This paper describes an iterative classification algorithm that classifies passenger boardings into two categories; transfer journeys and single journeys. The dataset used was from an urban public transport operator with a large fleet (over 1000 buses) and data of 48 million magnetic strip card boardings from 1998 and 1999. This paper details the process involved in the initial development of the iterative classification algorithm, the analysis of transfer node identification matrices, waiting time information charts and spatial first/second boarding matrices. When the algorithm is applied to the dataset it provides transport planners with valuable information with regard to passenger boardings, transfers and waiting times which can assist them in transport planning and policymaking. The purpose of this paper is to describe the automatic generation of a new data attribute that cannot be derived directly and therefore increases the future utilization of the dataset. The paper presents various analyses based on the extended and enriched database to illustrate this point.