Research And Application Of Railway Passenger Co-travel Prediction Method Based On Graph Pattern Association Analysis

With the continuous development of railway transportation technology and the gradual improvement of the railway network,the number of people travel by rail has been increasing,and the behavior of co-travel has become more and more obvious,e.g.,traveling with colleagues,traveling with friends.This co-travel relationship among individual travelers establishes a co-travel network over time.Predicting the potential links in this network can not only help the passenger transport,tourism and other related industries to provide personalized services and recommendations for co-travelers,thus improve the passengers’ satisfaction.It can also be used to support decision-making in the passenger transportation industry.Based on the information above,this paper proposes to use the classical data mining method association analysis to predict the co-travel relationship among railway travelers from the perspective of co-travel network.For this purpose,the paper firstly abstracts the prediction of co-travel relationship in the co-travel network into a link prediction problem,and studies the technology about graph pattern association analysis,including graph pattern association rules(GPARs)mining,graph pattern association rule evaluation and graph pattern association analysis.The experimental results show that our method is efficient and accurate in link prediction.Then this paper proposes co-travel graph pattern association rules(x-CGPARs and y-CGPARs),by extending GPARs based on the characteristics of railway travelers’ co-travel behaviors,which x-CGPARs will be used for predicting new co-travel relationships and yCGPARs will be used for predicting changes in existing co-travel relationships,and applies them to predict co-travel relationships in railway co-travel network.The experimental results show that the average accuracy rate of the method proposed in this paper is 0.797,while the accuracy of the traditional link prediction method Jaccard only has 0.390.Although the recall rate of Jaccard is slightly higher than the method proposed in this paper,the y-CGPARs can predict the changes in existing co-travel relationships,which cannot be achieved by the traditional link prediction method.Therefore,the method proposed in this paper can accurately and comprehensively predict the potential co-travel relationships in the railway co-travel network.Finally,this paper analyzes the application scenarios of the method proposed in this paper from four aspects,which are accurate travel recommendations for railway passengers,advance identification of key passengers,advance formulation of seat change scheme and advance warning of partnership crimes,and further clarifies the application value and management significance of the method.