Clustering as an approach for creating data-driven perspectives on air travel itineraries

Abstract

This paper proposes to create data-driven perspectives on airline ticketing data by defining groups of air travel itineraries independently of geographical or temporal attributes. We demonstrate the approach by analysing an extensive empirical data set featuring ticketing data from several carriers as collected by the Airlines Reporting Corporation. The analysis compares five cluster quality indicators to evaluate effects of pre-processing and clustering decisions. Benchmarking the results against a more traditional geographical grouping demonstrates the potential for data-driven analysis. Finally, we highlight ways in which the resulting cluster perspectives could support demand forecasting, performance evaluation, and analyst interventions.