A Multi-level Consensus Clustering Framework for Customer Choice Modelling in Travel Industry

Abstract

In the travel industry context, customer segmentation, that is the clustering of travelers to distinguish segments of customers with similar needs and desires, is a major issue for improving the personalization of recommendations in flight search queries. Indeed, when booking travel itineraries, different customers purchase tickets according to different criteria, like price, duration of flight, lay-over time, etc. However, clustering algorithm application is a challenging task because of two central issues inherent to the unsupervised nature of the grouping of instances: The choice of the clustering algorithm and parameterization and the evaluation of the resulting clusters of instances. Essentially, each clustering algorithm and evaluation measure relies on an assumption of the distribution model of the instances in the data space. The relevance of the resulting clustering mainly depends to which extent they are adapted to the analyzed data space properties. We present a Multi-level Consensus Clustering framework that combines the results of several clustering algorithmic configurations to generate a multi-level consensus clustering solution in which each cluster represents an agreement between the different clustering results. Relevant agreements are identified using a closed sets-based approach and represented in a hierarchical representation providing the end-user a representation of the consensus cluster construction process and their inclusion relationships. We show how this framework developed for Customer Choice Modeling in travel context can provide a better segmentation and refine the customer segments by identifying relevant sub-segments represented as sub-clusters in the hierarchical representation, and we present the technical and scientific challenges posed by the approach.