Research On Prediction Method Of Subsequent Itinerary Of Passengers In Unfinished Journey

In recent years,the intelligent service of civil aviation passengers has always been the main problem in the development and research of civil aviation industry.In the whole journey of civil aviation passenger,it is particularly important to provide intelligent services for the special group of passengers whose subsequent itinerary is not scheduled,these passengers are likely to have the potential demand for tickets or travel in the near future,and the demand for services will be more intense and concentrated.However,due to the sparse travel data and irregular travel intervals of civil aviation passenger,and the travel of passengers is affected by various factors,which is difficult to be predicted timely and accurately.If we can correctly locate these passengers,dig out the travel characteristics of these passengers with strong travel demand in the near future,and accurately predict the subsequent travel destinations of these passengers,it will help the development of intelligent services for civil aviation passengers,promote the overall industry revenue,and improve passenger satisfaction with services.First of all,it is the basis of the whole research work to correctly locate the passengers whose subsequent itinerary is not scheduled.In this study,a classification model of passenger current itinerary state based on graph neural network(IC-GNN)is proposed by studying the complete historical travel data.The model uses graph neural network to model all the complete historical itinerary of passengers,and emphasizes the starting city and current city through an attention module.Then it designs some artificial features which are closely related to the travel mode of passengers to supplement and output the current itinerary state of passengers.Secondly,the subsequent travel prediction of this kind of passengers is the core of this research.In this thesis,a graph structural multi-attention prediction model(GMAN)is proposed to predict the subsequent itinerary of passengers.In this model,the graph structure will be used to process the data of passenger travel sequence.Firstly,the characteristics of each city in the passenger travel sequence are described by the node representation of graph neural network.In addition,the global route network is introduced according to all the passenger travel data,and the city node representation which integrates the global route is obtained through the network representation learning,which is combined with the city node representation obtained by the graph neural network.Then,a multi attention network is introduced to capture the different dependencies among the nodes,and the importance of the current city nodes is considered to obtain the travel sequence representation of passengers.Finally,the probability distribution of the next travel city is obtained through the prediction module to predict the subsequent travel city of passengers.Finally,the passenger current itinerary state judgment model IC-GNN and the subsequent itinerary prediction model GMAN proposed in this research are tested on the real travel data set.Compared with many benchmark methods,the models in this paper are better than other methods.The experimental results show that the method proposed in this thesis can effectively identify the passengers whose subsequent itinerary is not scheduled,and accurately predict their subsequent itinerary.