Detecting weak public transport connections from cellphone and public transport data

Securing public transportation ridership is critical for developing a sustainable urban future. However, many modern and growing cities are facing declines in public transport usage. Existing systems for analyzing and identifying weaknesses in public transport connections face major limitations. In cities, origin-destination (OD) matrices--which measure the flow of people between different geographical regions--are often generated using household surveys, which are time consuming and lack spatial and temporal accuracy. Focus in more recent research has been drawn towards using cellphones to overcome these limitations. In this article, we show that urban mobility patterns and transport mode choices can be derived from cellphone call detail records coupled with public transport data recorded from smart cards. Specifically, we present new methods to determine the spatial and temporal variability of public and private transportation usage and transport mode preferences across the dense, metropolitan city-state of Singapore. The cellphone dataset consists of location data of 3.4 million anonymized users of SingTel, Singapore's largest telecommunications company with a market share of 45.3%. The data were recorded during a two-month period from mid-March to mid-May 2011. A call detail record (CDR) includes the location (spatial resolution of 400 m) of the cell tower each cellphone connects to and was created in the case of following network events: • a phone call was initiated or received (at the beginning and at the end of the call). • a short message was sent or received. • the cellphone user accessed the data network (for example, to open a website or retrieve emails). By applying a clustering detection algorithm to these call detail records, we detect individual trips and extrapolate the overall mobility of people between the 55 administrative districts of Singapore (taking into account the market share of SingTel and the cellphone penetration of 144 %). The mode share of private transport usage is then derived by computing the difference between the estimated overall mobility and the traces of 4.4 million public transportation smart card users during the same time period. We validate out data mining approach using the results from Singapore's Household Interview Travel Survey (HITS): Our results revealed that there are 3.5 million (HITS: 3.5 million) inter-district trips by public transport and 4.3 million (HITS: 4.4 million) inter-district trips by private transport (including taxis). Private transport usage dominates in regions without access to a subway line (see Figure 1). Along with classifying which transportation connections are weak or underserved---where people prefer to take private rather than public transport---the analysis shows that the mode share of public transport increases from 38% in the morning to 44% around mid-day and 52% in the evening. The value of deriving such patterns using cellphone call detail records have important implications not only for urban and transportation planning, but also for other domains such as disease control in cities. As humans serve as the primary and secondary vectors of many infectious diseases, understanding from where people arrive and depart and by which transportation modes people are traveling, we have the potential to model how and where diseases might be spreading and from where they might originate.