Optimization of ticket purchasing strategy based on machine learning

With the development of the aviation industry and the improvement of people's living standard, more and more people choose aircraft as their way of travel, but the airline adjusts the price according to the revenue management in real time. The purpose of this paper is to design different decision-making tools from the customer's perspective, and to provide customers with the relevant information needed to determine when to purchase a ticket properly. This paper models data using random forest algorithm and builds decision-making tools, taking into account the price fluctuation and amplitude of fluctuation, and suggests the customers when to buy their tickets and when to postpone the purchase. As an application, the paper analyzes air prices for routes from PEK to HET, CAN, SHA, DLC and CSX over the course of two months, and takes into account advance-purchase ticket sales of up to 40 days. By comparing and analyzing the forecasting performances of each model, the experimental results show that random forest model is more reliable in the condition of adding new features. Our study suggests that the model can serve as a basis for customers to make purchase decisions, so as to achieve the purpose of saving money for customers.