关键词: font-size:15.04px, ">航空运输, 特征矩阵, 门控循环单元, 图卷积神经网络, 航站楼旅客流预测

Abstract:

Aiming at the problem that it is difficult to accurately predict the spatial distribution of passengers in the terminal building over time, a prediction method of the spatiotemporal distribution of passenger flow is proposed based on graph convolutional neural network. Firstly, the time distribution matrix of passenger flow and regional adjacency matrix within each time slice is constructed to obtain the spatial feature matrix. The characteristic matrix of passenger flow in terminal building is extracted combining with the input gated recurrent unit, and performs linear integration with reference to the characteristic matrix of passenger flow caused by the change of the boarding gate.

Then, the final characteristic matrix of passenger flow of the space area is obtained. Finally, validation analysis is conducted based on actual data from a domestic hub airport. The results show that the proposed method improves the goodness of fit by 7% to 10% compared to the traditional models, providing effective support for the strategic optimization of passenger operation resource guarantee and comprehensive transportation capacity in terminal building.

Key words: font-size:15.04px, ">air transport, characteristic matrix, gated recurrent unit, graph convolutional neural network, passenger flow prediction in terminal building