Flight delay causality: Machine learning technique in conjunction with random parameter statistical analysis

Seyedmirsajad Mokhtarimousavi

doi:10.1016/j.ijtst.2022.01.007

2023-03-01

Seyedmirsajad Mokhtarimousavi *

Armin Mehrabi

DOI：10.1016/j.ijtst.2022.01.007

原文链接

查看原文

摘要

En 中文

The consequences of flight delay can significantly impact airports' on-time performance and airline operations, which have a strong positive correlation with passenger satisfaction. Thus, an accurate investigation of the variables that cause delays is of main importance in decision-making processes. Although statistical models have been traditionally used in flight delay analysis, the presence of unobserved heterogeneity in flight data has been less discussed. This study carried out an empirical analysis to investigate the potential unobserved heterogeneity and the impact of significant variables on flight delay using two modeling approaches. First, preliminary insight into potential significant variables was obtained through a random parameter logit model (also known as the mixed logit model). Then, a Support Vector Machines (SVM) model trained by the Artificial Bee Colony (ABC) algorithm, was employed to explore the non-linear relationship between flight delay outcomes and causal factors. The data-driven analysis was conducted using three-month flight arrival data from Miami International Airport (MIA). A variable impact analysis was also conducted considering the black-box characteristic of the SVM and compared to the effects of variables indented through the random parameter logit modeling framework. While a large unobserved heterogeneity was observed, the impacts of various explanatory variables were examined in terms of flight departure performance, geographical specification of the origin airport, day of month and day of week of the flight, cause of delay, and gate information. The comprehensive assessment of the contributing factors proposed in this study provides invaluable insights into flight delay modeling and analysis. & COPY; 2022 Tongji University and Tongji University Press. Publishing Services by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/ licenses/by-nc-nd/4.0/).

Keyword:

Flight delay

Air-traffic management

Random parameter logit model

Machine learning

Support Vector Machines