Dynamic Differential Pricing of High-Speed Railway Based on Improved GBDT Train Classification and Bootstrap Time Node Determination

Abstract

This paper proposes a differential pricing method and a strategy for determining the time node based on the multi-dimensional characteristics of trains by Markov stochastic process, which focus on the study of dynamic differential pricing of the same class of seats with non-parallel trains with different operating characteristics at the same origin-destination. Firstly, an improved machine learning Boosting model Gradient Boosting Decision Tree(GBDT) is proposed to optimize the price discrimination grouping. Secondly, a dynamic simulation mechanism of passengers’ selection under differential pricing is constructed, which breaks through the traditional measurement method to measure the passengers’ sensitivity to price changes. Thirdly, a new method, which uses Bootstrap to get the dynamic feedback for the adaptability of passengers under a differential pricing system, is proposed to determine the iteration nodes of differential pricing. Finally, based on the Markov property of railway passengers’ selection, the paper breaks through the traditional static pricing mechanism and constructs a dynamic differential pricing solution framework. Taking the Beijing-Shanghai high-speed trains from Beijing South to Shanghai Hongqiao as an example, by comparing with the existing dynamic differential pricing system, it’s proved that the dynamic differential pricing system proposed in this paper can balance the occupancy rate between trains, suppress the fluctuation of occupancy rate, determine the updating node of differential pricing, etc., and hence, is more reasonable and effective than the current system.