Spatiotemporal Hotspot Pattern Analysis of Air Travel Passengers with Long Short-Term Memory Networks

The surge in air travel and diverse transportation methods have intensified the challenges faced by airport landside traffic management. Conventional solutions, such as expanding terminal roadways, have become costly and inefficient in meeting the demands of intelligent traffic management. Leveraging the wealth of data, including GPS, deep learning algorithms offer a more effective approach to identify spatiotemporal patterns in air passenger travel and forecast hotspot areas. By proposing a Long Short-Term Memory (LSTM)-based algorithm for predicting these hotspots and enhancing its accuracy through data fusion, this paper empowers airport authorities to proactively manage traffic congestion. The experiment showcases the algorithm’s superiority over baseline models, particularly in comparison to curve-fitting approaches. The combination of highly correlated heterogeneous data, like landside traffic flow and flight data, augments the accuracy of deep learning models. However, this predictive accuracy diminishes when data correlations are low. This methodology equips airport managers with insights into future traffic trends, serving as valuable reference data for future airport planning, thereby advancing airport management’s digital evolution.