ABSTRACT
This paper combines the k-means clustering method in combination with PCA and the system dynamic modeling approach to derive a better insight into the behavior of airline profitability during the time span of 1995 until 2020. The model includes various explanatory variables that capture different aspects of airline economic and operational metrics, whose fluctuations may affect the airline profitability. By forecasting these exogenous variables, the system dynamic model is used to predict airline profitability through 2025 and answer the question of whether the US airline industry will return to its pre-COVID 19 pandemic state. The latter research question can be agreed with, as the effect of introducing a fourth dimension derived from Principal Component Analysis (PCA) to sufficiently cover the variation within the dataset during the years of COVID-19 pandemic diminishes towards the end of the forecast period. Furthermore, the key measures from PCA imply that under the assumption of continuous growth and a non-exogenous shock, future years will not cluster in past years. The six different clusters from 2019 to 2025 showed how the system stays in a certain state for a few years and then drifts further to a new state. There are only a few variables that change to transfer from one cluster to the next. © 2022 The Authors
ABSTRACT
This paper combines the k-means clustering method in combination with PCA and the system dynamic modeling approach to derive a better insight into the behavior of airline profitability during the time span of 1995 until 2020. The model includes various explanatory variables that capture different aspects of airline economic and operational metrics, whose fluctuations may affect the airline profitability. By forecasting these exogenous variables, the system dynamic model is used to predict airline profitability through 2025 and answer the question of whether the US airline industry will return to its pre-COVID 19 pandemic state. The latter research question can be agreed with, as the effect of introducing a fourth dimension derived from Principal Component Analysis (PCA) to sufficiently cover the variation within the dataset during the years of COVID-19 pandemic diminishes towards the end of the forecast period. Furthermore, the key measures from PCA imply that under the assumption of continuous growth and a non-exogenous shock, future years will not cluster in past years. The six different clusters from 2019 to 2025 showed how the system stays in a certain state for a few years and then drifts further to a new state. There are only a few variables that change to transfer from one cluster to the next.