Performance analytics and data centric engineering
The users and suppliers of transport services generate an enormous amount of digital signs every time they buy a ticket, enter a station, close a train door or pass a trackside signal, for example. Planners, policy makers as well as researchers can learn a lot from these digital records, as the entire pattern of daily passenger movements and technological efficiency metrics are encapsulated in the resulting large-scale datasets. Our experience suggests, however, that transport agencies cannot fully utilise the new datasets that recently installed electronic fare payment or signally signalling systems, among many others, have made available. The primal reason why rich datasets are often stored for a while and then deleted without profitable information extraction is the absence of immediately available methods for specific types of quantitative analyses.
In the recent years, smart card and other automated data based research topics have become very popular in the academic sphere, thus resulting in a wide variety of research outcomes of mixed quality. Our research group contributes to this emerging field with (i) state-of-the-art econometric and causal inference methods that go beyond what black box data mining techniques can reveal, and (ii) the utilisation of over 20 years of experience we gained through benchmarking activities and collaborations with public transport operators all over the world. Our aim is to come up with research questions that are relevant for the industry and turn large-scale transport datasets into impactful results that the public transport sector can utilise in everyday decision making.
Hörcher, D., Graham, D. J., & Anderson, R. J. (2017). Crowding cost estimation with large scale smart card and vehicle location data. Transportation Research Part B: Methodological, 95, 105-125.
Ongoing Research Projects
Using smart card data to analyse the disruption impact on urban metro systems
Incidents occur regularly on urban metro systems. It is common to encounter a disruption, especially for those systems operated for over a century, such as London Underground and New York Subway. The occurrence of disruptions is likely to cause delays and disorders in the metro operation. Therefore, it is important to quantify the influence of metro disruptions.
This study aims to use smart card data to analyse the impacts of disruptions on London Underground. First, we estimate the causal effects of metro disruptions on travel demand and average journey time at multiple aggregation levels. Then, second-stage regression models are built to find determinants of estimated disruption impacts. The research results will help metro operators prepare better recovery plans for future service interruptions.
Authors: Nan Zhang, Daniel J. Graham, Daniel Hörcher, José M. Carbo
Current status: Under development
A causal analysis of the impact of differential pricing on commuter trip scheduling in Hong Kong
To address overcrowding problems in public transit services, transit operators often adopt demand management measures aimed at spreading peak hour demand, including
'soft policies' in the form of monetary incentives to in uence passenger behaviour. Identifying the causal impact of such policies using a traditional travel survey is challenging due to the resource-intensive nature of such surveys. Smart card data offers the possibility of observing highly disaggregate patterns of passenger volumes continuously over time. It provides a level of granularity that allows us to estimate causal behavioural responses at a highly disaggregate level and then link these to observed contextual characteristics. Obtaining causal estimates from smart card data, however, requires application of appropriate methods, which can allow the research to disentangle the direct impact of monetary incentives on regular commuter's trip scheduling behaviour from other potential in uences on demand. To this end, this paper quantifies the causal impact of dierential pricing on trip-scheduling by regular commuters on the Mass Transit Railway (MTR) in Hong Kong. We use the difference-in-difference method to estimate the causal effect of the Early Bird Discount (EBD) on commuter trip-scheduling. We find significant but small effect from introduction of the EBD in terms of earlier departure times. In a second-stage regression we identify travel costs, crowding and work-schedule exibility as the key determinants of commuter responsiveness to the policy. Our analysis suggests that higher income commuters are more responsive, so the policy might be regressive.
Authors: Anupriya, Daniel J. Graham, Daniel Hörcher, Richard J. Anderson
Current status: Under review
Decomposing journey time variance on urban metro systems via semiparametric mixed methods
To attract and retain ridership, metro transit operators must deliver reliable and predictable service. In recent years with the availability automated data, it has become easier for operators to measure journey time reliability; however, there is limited understanding of why journey times vary or how they can be improved. In this paper, we present a semiparametric regression analysis to determine the underlying drivers of journey time variance on metros, taking the London Underground as a case study. We merge train location and passenger trip data sets to decompose journey times into three constituent parts: access times as passengers enter the system, on train times, and egress times as passengers exit at their destinations. Within each component, we derive elasticity estimates of journey times with respect to service supply and demand factors, including operational and physical characteristics of a metro system as well as passenger demand and individual-specific travel characteristics. We quantify and rank which parts of the system perform comparatively best and worst in terms of journey time variance at a station, route, and line level. Overall, the outputs of this analysis can be directly applied by operators to identify the causes of journey time variance, and undertake comparative benchmarking of stations and lines within systems
Authors: Ramandeep Singh, Daniel Hörcher, Daniel J. Graham, Richard J. Anderson
Current status: Under review