Machine Learning Approach to Analyze Airport Boarding Strategies Based on Passenger’ Behavior

• Marco Aurélio Gehlen
• Giovanna Miceli Ronzani Borille

Aeronautics Institute of Technology (ITA), Department of Civil Aeronautical Engineering

Event

XXI SITRAER – Fortaleza, CE, Brasil, 2024

Volume

Anais do XXI Simpósio de Transporte Aéreo – SITRAER 2024 · ISBN 978-65-272-1008-5 · DOI 10.29327/xxi-sitraer-2024

Pages

417–426

Publisher

Even3 Publicações

Editors

Francisco Heber Lacerda de Oliveira, Fabiano Rocha, Giovanna Miceli Ronzani Borille, Evandro José da Silva, Mauro Caetano

Abstract

Boarding is on the critical path for the turnaround time. It caused delays that reached almost $30billion in the pre-pandemic period according to FAA. The literature around airport boarding discusses which are the boarding strategies that minimizes boarding time, focused on pre-defined strategies that usually do not consider the context of passenger’ behavior (agility, hand-on bags, if alone or in groups) for a specific flight. Because of that, there is no consensus in the literature about the best way to reduce boarding time and consequentially to increase perceived Level of Service (LoS) especially considering passenger’ behavior. Aiming to contribute to this challenge, this study proposes using passenger behavior variables (queue interferences, bag distribution, priority rates and group travelling) to model boarding time for different boarding strategies. For that, this study combines two different techniques that can be replicated for other contexts: 1) A simulation of different boarding strategies based on a screening Design of experiments (DOE) to estimate boarding time for different passenger’s behavior contexts in a A230 model 2) Five machine learning techniques applied on the simulation dataset. As results, a model that can be used to predict boarding time for a given context of passenger behavior is discussed which can assist airlines in the decision making process to set boarding strategies based on a data-driven approach, which helps to avoid delays and to increase passenger satisfaction. Regarding the literature contribution, this study takes steps to fulfill the gap around setting boarding strategies to minimize boarding time by providing a way to define them based on each context of passenger behavior. It was also found that between the simulated boarding strategies (random, back to front and Steffen’s) the one that provides the lower boarding time is Steffen’s with an average of 12+-0.02 minutes to board 180 passengers.

Keywords: Boarding Strategy; Boarding Simulation; Machine Learning