Digital click stream data for airline seat sale prediction using GBT

Abstract

Revenue Management is important for every airline business and the seat is the main product of an airline. The purpose of the revenue management is to maximize the revenue of each airline routes based on demand. This demand, however, depends on factors such as historical demand, seasonality, seat pricing based on purchase lead days, competitors pricing and customer behaviour. Prediction of passenger demand helps to forecast revenue on future flights and thus allow the airline to generate optimal prices for the corresponding flights. Therefore, minimizing the prediction error constitute the most crucial goal of good revenue management. In this paper, A GBT based model has been proposed for airline seat sale prediction to optimize the revenue. To optimize the prediction accuracy, an analytic dataset has been developed by combining digital attributes and traditional operational and transactional attributes. This paper will also highlight an efficient data extraction and processing pipeline have been proposed to aggregate a large volume of unstructured data from various data sources. The empirical findings suggested applying GBT on transformed dataset can predict seal sales for 30 days ahead with accuracy of 93%.