A method for identifying connected flights in aviation schedules

Abstract

This paper describes a method of grouping flights in airline schedules into tail-connected itineraries. The purpose is to improve the realism of large-scale aviation simulations by allowing them to account for propagated delay, the source of about a third of all delays. The approach presented is that of probabilistic classification with supervised learning. Training data comes from the Airline Service Quality Performance Metrics (ASQP) database (www.bts.org). This data consists of scheduled arrival and departure times, aircraft tail numbers, carrier names, and aircraft types (i.e., Boeing-737) for about a third of all scheduled flights. The classification method described here is by necessity extendable to airports and aircraft types that are not in ASQP.