Modeling and Forecasting the Customer Activity for an European Travel Website

Abstract

How can predictive analysis of customer data be used to identity the decrease or increase of customer activity over time during the product search and the booking process? Another question that arises in this context is how many data is necessary to carry out a reliable analysis?

We present a case study from an European travel website based on a session history of 15 months only here only a small part of the customers are known to the website and where a huge number of users regularly delete their browser histories.

We focus on the analysis of the users' web traffic while visiting the booking platform. Based on a classification algorithm we identify the non-contractual customers who are mostly anonymous to the website, and investigate the short-term customer activity with respect to customer churn on a more microscopic level.

Keywords: predictive analysis, booking platform, churn, e-commerce, neural networks, random forest

JEL Classification: C52, C53, C81