Research Suggests Transit Agencies Can Improve Customer Satisfaction with Available Data

Bus transit agencies often provide their riders with estimated bus arrival times, based on real-time transit information, in order to help them make decisions, and improve their perceptions about the bus system’s reliability. A drawback has emerged, however: travelers believe that these estimates are accurate to the minute, and when they aren’t that precise, the result can be the kind of negative experience that reduces ridership.

Additionally, passenger occupancy can be a useful type of information for bus riders to have as they plan their trips, so they will be likely to know if a seat will be available on the bus when it arrives, or even if there will be space to board it at all; this type of information is sometimes provided to train passengers, where it has proven helpful, and it is already employed in a handful of bus transit systems.

A new study by the Mineta National Transit Research Consortium, Estimating Uncertainty of Bus Arrival Times and Passenger Occupancies, determines that accuracy for arrival time may be improved using different modalities than those commonly applied by transit agencies. It also explores techniques for estimating occupancies of the buses when they are still a number of stops away from any given point. In both cases an emphasis was on quantifying the level of uncertainty in the estimates. It is also suggested that bus passengers be given a way to gauge the confidence interval of these estimates, even as they are made as accurate as possible. The study was conducted by Vikash V. Gayah, PhD, who conducted the research with Zhengyao Yu and Jonathan S. Wood.

Says Dr. Gayah: “In an environment that is information-rich and in which transit users seek the most high-quality information about the current state of the transit network, providing these results to passengers might help to improve their decision-making and increase their confidence in the reliability of real-time transit information systems.”

The estimation of travel times was approached by comparing linear regression models to the newer technique of accelerated failure time (AFT) survival models. AFT survival models are used to predict the time remaining until an event occurs. Commonly used to model time-to-failure of infrastructure elements, they are here applied to a different event: the arrival of a bus at a downstream stop. The two techniques were used to create estimates that were then compared to actual times on a bus transit system. For estimation of passenger occupancies, linear regression and count regression models were considered. Each approach was then compared to actual bus occupancies. Quantile regression models are also proposed as a way of estimating confidence intervals, in order to correctly evaluate the uncertainty level for each model.

Dr. Gayah points out that “These models can also benefit transit service providers. For example, models of passenger occupancy can predict when buses are expected to be full so that additional capacity can be provided in real time. Travel time uncertainty can also be used to optimize staffing decisions and plan driver shift changes.”

For a free, no-registration download, go to http://transweb.sjsu.edu/project/1246.html