Evaluating Models for Estimating Schedule Deviation in Public Transit

According to a survey performed in 2013 on public transportation riders, there is an increasing number of users relying solely on real-time information for route schedules. As this trend continues, researchers and transit providers have developed algorithms to better predict schedule deviations and arrival time estimations. However, several number of these approaches have not been thoroughly documented and evaluated, limiting the reproducibility of results across publications. This paper identifies these limitations and proposes evaluation methods that can be more easily compared. To showcase these approaches, we evaluated three different predicting models used for schedule deviation across two cities.