Predictive Analytics for No-Shows and Cancellations for Paratransit Operations

“Llame y Viaje” (Call and Travel): Puerto Rico’s Autoridad Metropolitana de Autobuses (AMA) operates a total of 43 busses equipped for passengers with physical or mental handicaps.

Publicly financed paratransit systems are a critically important social service for individuals with disabilities who do not have access or cannot afford alternative modes of transportation. Unfortunately, paratransit services are also one of the most expensive services provided by transit agencies.

 By federal law, the fare that a transit agency can charge for a paratransit trip cannot exceed twice the fare charged for a comparable trip on its fixed-route bus system, and as a consequence paratransit systems need to be heavily subsidized to cover operational costs.

Naturally, transit agencies have attempted to implement various measures to improve the overall efficiency of their operations and reduce operational expenses. Among these measures are policies aimed at reducing no-shows and late cancellations (i.e., two situations in which a scheduled trip is not performed). No-shows and late cancellations are events that waste transit agencies’ resources, as well as degrade a system’s productivity by preventing other users from utilizing supplied service slots.

Given that no-shows and late cancellations are part of the normal operations of paratransit systems, the analysis and prediction of their occurrence should be a formal component of operational and planning models for paratransit systems in order to improve productivity and decrease operating costs. However, little to no attention has been given to understanding and predicting no-shows or late cancellations for paratransit operations. In addition, the uncertainty introduced by no-shows and late cancellations has not been formally incorporated in the crucial operations processes of designing vehicle routes and booking trip requests in paratransit systems.

The objective of this research project is threefold:

  • to develop a classification methodology to predict no-shows and cancellations of trips in paratransit systems,
  • to incorporate the classification model predictions into trip booking and routing models for paratransit operations (including novel overbooking problems in the context of paratransit), and
  • to evaluate the value of the classification models’ predictions in paratransit planning and operations.

Researchers: Jee Eun Kang (University at Buffalo) and Daniel Rodriguez (University of Puerto Rico at Mayagüez Campus).

Partners: The researchers will collaborate with Puerto Rico's Metropolitan Bus Authority (AMA, by its acronym in Spanish) to develop and apply the data analytic techniques proposed in this project..

Data: From Puerto Rico's largest publicly operated Paratransit System: 'Llame y Viaje" (Call & Travel).