Clustering of airport cities and cluster dynamic for the air passenger demand forecasting model based on a socio-economic scenario

Abstract

This study presents methods of grouping cities into clusters by their socio-economic indicators and tracing changes in the content of cluster within a socio-economic scenario. For cities’ grouping, three main clustering approaches have been analyzed: hierarchical, exclusive and probabilistic clustering. Analyzing advantages and disadvantages of these approaches, probabilistic clustering of normal mixture has been chosen to separate cities from the air passenger demand (APD) forecasting model. Three parameters, a city’s GDP, population and GDP per capita, have been used for clustering. Utilizing these parameters and based on special metrics, separation of cities into 9 clusters has been chosen. Furthermore, this study introduces the “cluster dynamic”. The cluster dynamic defines how cities are allocated to the various clusters at a given point in time within a socio-economic scenario.