Short-term Passenger Flow Forecasting of the Airport Based on Deep Learning Spatial-temporal Network

Abstract

In recent years, with the rapid growth of airport passenger flow, airport services such as security inspection, emergency response, check-in, baggage tracking are facing tremendous pressure. Being able to make relatively accurate predictions of short-time passenger flow in airport terminals is an important basic guarantee for improving service quality, enhancing operational efficiency, and rationalizing resource allocation. In this paper, we establish a multi-gate short-time passenger flow prediction model ST- LSTM based on deep spatial-temporal learning, which integrates convolutional neural network (CNN) and long short-term memory (LSTM) to improve short-time passenger flow prediction accuracy. Based on the actual passenger flow data of 2 million departing passengers at Guangzhou BAIYUN International Airport, according to the number of passengers connected to Wi-Fi AP (Access Point), flight schedules, gates area, Wi-Fi access point location characteristics, etc. Through comparison with the HA, ARIMA, GBDT, LSTM, it is proved that the ST-LSTM model can more effectively predict the short-term passenger flow of the airport, which provides crucial decisions for the dynamic allocation and optimization of resources in the boarding gates, gives guiding significance to actuality.