Airline Customer Segmentation based on Complex Behavioral Approach using K-Mode and XG-Boost Algorithm

Abstract

Passenger disappointment with quality of services is one of the key component affecting the revenue deprivation of the airline industry. Now the airline industry has realized that traditional customer segmentation based on demographics does not reflect customer behavior. Airline is one of the global industries in which customer expectations change very rapidly. Dealing with those expectations in a highly competitive market means airlines must re-formulate their customer segmentation process, from social demography to more composite behavioral approach that consider the entire travel experience of the customer. To overcome above limitation, this study uses passenger booking GDS data having attributes such as user origin, destination, flight class, price, travel search category, trip span, time of fly, number of passengers, trip break and is round trip in order to predict the type of traveler I.e. whether a user is a family traveler, group traveler, business traveler and solo traveler. To find the pattern in the data this study illustrates the idea of applying K-Mode clustering which calculates the most optimal clusters within the data points. Then to predict the class of the traveler XG Boost algorithm is applied and for explaining the outcome of any machine learning model. Shapley Additive Explanations value analysis is used.