My research interests earlier this year began in understanding what drives certain behaviour in the urban environment – specifically in what changes to transport systems do in affecting people’s choice to use public transit. It’s been a long haul, but today marks the completion of this project, titled ‘The use of automatically-collected transport data in the spatiotemporal analysis and visualisation of policy change: a case study of the San Francisco Municipal Railway’. Here I will provide a bit of an overview for those who might be interested.
There are a wide array of technologies within our urban environment that are constantly collecting data. Many of these systems are still in relative infancy, however for transit systems have been among the earliest to be implemented and mature. While many of these sensors, particularly in transit, are used for real-time decision making, there is an increasing interest in the use of this data for retrospective analyses. Studies over these long periods provide improved means in understanding changes in behaviour, as they can potentially be provided with enough data to achieve statistically sound results.
The dissertation focuses on utilizing data that has been collected throughout the public transport network of the City of San Francisco. San Francisco are innovative in their approach towards collecting data to improve their public transit network. In particular, they have been collecting detailed data of their bus network over the past five years. This data includes automatic-vehicle-location (AVL) and automatic-passenger-counting (APC) data. This vehicle location data, which many people are exposed to every day through ‘next bus’ information in signs or on their mobile devices can provide very detailed information on the transport system.
When archived over large periods of time, it can draw a picture of how the systems has been performing and how it has changed, from a stop-level, to route-level and system-wide perspective. There are many performance variables that can be deduced from these samples of vehicle location – including the how often the bus arrives on time, the speed of the bus and the waiting time for passengers. While this gives us an indication of the vehicular performance of the fleet, we can also ascertain an understanding of passenger experience, and from their ridership changes, how this experience affected their behaviour.
This is achieved through APC data, which people may be less aware of. This data is collected through laser beams above the door of the bus which perform count calculations of how many people entered and exited the vehicle based on how the beams were broken. Counting data could also be used from information sources such as fare-collection boxes, where you touch on your Oyster Card or Myki.
The dissertation uses these large datasets to assess how both people and vehicles changed over time throughout the city. During the research, it was decided to focus on one particular policy where a large change was hypothesized to have occurred in the system. A number of changes were identified, while a number of variables also appeared inelastic to the policy change.
As part of this dissertation, two specific visualisation and analysis techniques were employed. Firstly, this involved the development of an interactive visualisation tool which enabled rapid exploratory analysis of the data. Secondly, two networks were created to represent the data – one where all consecutive bus stations were linked, and the other where all bus stops with a shared route were linked. With these networks, a number of graph-theoretic attributes were tested, weighted with the data at different time periods – including degree, betweenness centrality and efficiency.
Throughout the process it was found that the policy changes and effects were able to be portrayed in the automatically collected data, however not without its own challenges and potential inaccuracies based on its size and sampling methods. Further research in this field could involve utilising similar tools and methods in understanding at a smaller time scale, such as the system response after a tube strike or natural disaster. If you are interested, pieces of this work feature in the publication listed below and the full dissertation will possibly be available in several months after marking at CASA, UCL library.
Lock, O. 2014. The use of automatically-collected transport data in the spatiotemporal analysis and visualisation of policy change: a case study of the San Francisco Municipal Railway. Diss. University College London (UCL). London, United Kingdom.
Erhardt, G. D., Lock, O., Arcaute, E. & Batty, M. 2014. A Big Data Mashing Tool for Measuring Transit System Performance. The Big Data and Urban Informatics Workshop. Chicago, Illinois, USA.