Deriving Transit Performance Metrics from GTFS Data

Transit agencies devote extensive resources to producing General Transit Feed Specification (GTFS) Schedule and Realtime data to power trip planning applications. In representing scheduled and actual service characteristics, these data offer a theoretical off-label use to generate metrics of transit service performance. This project seeks to create and test a set of standardized protocols for deriving and visualizing these performance metrics from raw GTFS feeds. These protocols would be established in such a way as to enable transit agencies, planning organizations, transportation researchers, transit advocates, and community-based organizations to easily implement them on any transit system with available GTFS schedule and GTFS realtime feeds. Specific metrics would look at common transit issues tied to schedule deviation – from on-time performance to bus bunching – but at a more granular spatial and temporal level than ever before possible. This detail, literally at the stop and segment level, is designed to enable more effective transit planning and advocacy.

The research would first collect a multi-day sample of GTFS schedule and realtime data from one or more transit agencies. This information would serve as the core data for the entire project. These data would be stored in a relational database that enables spatial analysis, such as PostGIS. The research would first design appropriate cleaning and aggregating protocols to prepare the data for performance analysis. A web-based tool, likely using D3, would be designed to allow a user to interact with the data to generate and visualize the transit performance metrics. The tool would allow fine-grained filtering by location and time period to enable detailed analysis of transit performance. A key feature of this approach is to allow interactivity with the data.