A Data-Driven Agent-Based Simulator for Air Ticket Sales

In order to better design sales strategy, air companies or travel agents would predict sales of air tickets. In this paper, we propose an agent-based ticket sales simulator based on data analysis. The features of air tickets and passengers are extracted by analyzing real data. A Long Short-Term Memory recurrent neural network is used to forecast the daily customer search volume. Then a purchase decision tree is designed and embedded into the customer agent to simulate the decision process when a customer tries to find and buy an air ticket. Experimental results show that our prediction model achieves better prediction accuracy than three compared approaches. Moreover, through the simulation experiment on the historical real data, we obtain good simulation results, and verify the validity and practicability of our ticket sales simulator.