RT14 – Dr Zhenliang Ma – Harnessing data science in transit operations and planning
Researching Transit – Episode 14
Published: September 2020
Keywords: data science, data analysis, visualisation, data storytelling, public transport, transit, operations, planning, crowding
Dr Zhenliang Ma is a researcher and lecturer at Monash University and co-director of the graduate transport program jointly run by Monash University and Southeast University in China. Dr Ma moved to Monash after working at MIT, to join its interdisciplinary public research team.
This episode addresses the potential for data analytics to help transit agencies diagnose problems and identify opportunities to improve operations and customer satisfaction. Dr Ma provides some examples of problems that are suited to a data-driven solution, and some that aren’t.
“[Data analytics] is used to try to transform data into information to derive insights, and from those insights, make better decisions.”
He characterises three particular applications of data analysis to transit problems:
- Inferences problems, which leverage descriptive and diagnostic problems, which use travel data to understand system performance, and passenger decision making
- Prediction problems, which use predictive analysis to improve real time control of vehicles based on traffic conditions and disruption
- Long-term demand management problems, which use prescriptive analysis to test how users would respond to different incentives designed to change travel behaviour.
We discuss the application of prescriptive data analysis to address severe crowding in Hong Kong’s metro system (a similar solution to peak crowding is discussed in the Singapore context by Dr Waiyan Leong in Episode 7). This project sought to improve the payoff for demand management interventions by identifying users most likely to respond.
Dr Ma mentions a trial of a personalised incentive system for San Francisco’s Bay Area Transportation Authority (BART). The success rate of demand management improved for incentives targeted toward individuals rather than the station. Data analysis was used to understand behavioural responses to incentives, and to design the final demand management strategy to optimise success.
“The transportation system is very complex. By changing a small portion of the passengers behaviour, congestion will be solved”
Dr Ma defines three steps to tackle constrained public transport capacity during the COVID-19 pandemic. Highlighting this very deliberate approach to thinking about data science problems, he does so in the language of data science, proposing first to use data to describe usage patterns, diagnose problematic times, and predict what response might occur under different policy scenarios. Being deliberate in the way you approach the problem is key.
“We really need to think about how to represent our data, to tell the story or to understand the problem, and then we can develop new insights from that”
However, although data analysis is useful for exploring a problem, it cannot explain why. Data-driven solutions alone are not enough to understand why human make decisions.
“Data science is just one of the tools, out of the set of tool that we can use to solve transport problems”
What makes a great public transport data analyst? First, is an interest in data, and open and skeptical mind (prepared to challenge results). Skills in programming, statistics and visualisation will give the aspiring data analyst a toolbox for their work. Finally and most importantly, is domain knowledge.
Dr Ma recommends aspiring data analysts develop skills in programming (he recommends python), statistics, and visualisation tools for telling a story with the data so it is easily understood. He recommends the following resources:
- Machine Learning lecture series (open access), Andrew Ng, Stanford University, https://www.youtube.com/playlist?list=PLLssT5z_DsK-h9vYZkQkYNWcItqhlRJLN
- Artificial Intelligence Algorithms lecture series (open access), Professor Patrick Winston, Massachusetts Institute of Technology, https://www.youtube.com/playlist?list=PLUl4u3cNGP63gFHB6xb-kVBiQHYe_4hSi
- Find, collaborate and share simple (or complex) software code on Github: https://github.com/
- Kaggle (Machine Learning and Data Science Community): https://www.kaggle.com/
Read about the case study mentioned in today’s show in Dr Ma’s publications:
- Halvorsen, A., Koutsopoulos, H. N., Ma, Z., & Zhao, J. (2020). Demand management of congested public transport systems: a conceptual framework and application using smart card data. Transportation, 47, 2337–2365. doi: https://doi.org/10.1007/s11116-019-10017-7
- Ma, Z., & Koutsopoulos, H. N. (2019). Optimal design of promotion based demand management strategies in urban rail systems. Transportation Research Part C: Emerging Technologies, 109, 155-173. doi:https://doi.org/10.1016/j.trc.2019.10.008
- Ma, Z., Basu, A. A., Liu, T., & Koutsopoulos, H. N. (2019). Behavioral Response to Transit Demand Management Promotions: Sustainability and Implications for Optimal Promotion Design. Paper presented at the Transportation Research Board 98th Annual Meeting, Washington DC, United States.
Sing up for updates when we release shows: http://eepurl.com/g9tCdb
Music from this episode is from https://www.purple-planet.com
- Date September 27, 2020
- Tags Crowding, Fares, Hong Kong, Mass Transit, Podcast, Policy, Rail, Subway/Metro, USA