Research On Prediction Method Of Spatio-temporal Distribution For Arriving Passenger Flow In Airport Transportation Hub

With the increase of civil aviation passenger flow in recent years,the airport transportation hubs have experienced the situation of long queue time and low service quality for arriving passengers in many degrees so that the traditional methods to allocate capacity resource of airport transportation hubs have no longer fitted current status of civil aviation development.How to establish a stable and accurate short-term arriving passenger flow prediction method has become the key issue to achieve fine operation and management of airport transportation hub,the accuracy of prediction method directly affects the efficiency of traffic capacity resource allocation and travel experiences of passenger.This thesis compared and analyzed common short-term prediction methods,found that the prediction method based on graph convolutional network is more suitable for predicting the temporal and spatial distribution of arriving passenger flow in the airport transportation hub.Arriving passengers in airport transportation hub has strong purposes of movement,and the number of areas which passengers pass though in short time is large.So using traditional prediction methods based on graph convolution network has the defect of over-smoothing.This thesis designed an improved graph convolutional network which both set adjacency matrixes of inflow and outflow to extract the spatial features of the distribution of arriving passenger flow,and designed gated recurrent network is used to extract the temporal dependence of the spatial feature sequences and eventually constructed the short-term arriving passenger flow prediction model based on deep spatio-temporal graph convolutional network.Experiments showed that this model effectively alleviates the over-smoothing problem,and the prediction result has higher accuracy than the traditional prediction model.Since the arriving passenger flow is driven by the flight,this thesis combined the historical passenger flow distribution and historical flight arrivals,and designed a flight information correction module to correct the prediction results which further improves the prediction accuracy of the model.On the basis of obtaining the prediction results of the spatio-temporal distribution of arriving passenger flow of the airport transportation hub,this thesis selected similarity of internal factors,similarity of external structural,sample entropy and statistic complexity to measure the passenger flow situation,and used the improved K-Means method to cluster and analyze the various indicators.Based on the clustered results,this thesis showed that the method can accurately reflect the situation of arriving passenger flow distribution.