Research On Bus Line Selection Prediction Based On Traffic Card Big Data Platform

Urban public transportation,as the main mode of travel of the citizens,has the function of diverting dense crowds and plays an important role in alleviating traffic congestion.However,with the rapid development of the economy and the acceleration of the urbanization process,the number of floating population has increased rapidly,which has caused tremendous pressure on urban public transportation.At present,the contradiction between limited public transportation resources and huge travel demand has brought many problems to passengers and transportation authorities in planning bus routes.For example,some passenger lines have too many passenger flows,which brings extremely poor bus ride experience to passengers;the line shunt and load function are not balanced,which further aggravates the congestion level of bus lines.The key to solving these problems is to establish a correlation model between passengers and lines to predict the passengers’ choice of bus routes.Therefore,the bus route selection predi ction problem based on the traffic card big data platform is of great significance.Focusing on the research of this topic,the main work of this paper includes the following aspects.(1)In order to better establish the association model between passengers and bus lines,this paper builds a more comprehensive feature set.In view of the shortcomings of previous scholars on factors affecting passenger travel,comprehensive visual analysis of time,date,weather information and crowds.According to the results of visual analysis,the feature set of passenger and line characteristics,time characteristics and weather features are constructed,and the optimal feature subset is selected by combining filtering method and integrated model method.(2)This paper uses the eXtreme Gradient Boosting Model to predict passengers’ bus routes.First,the data used for modeling is obtained through Sqoop and web crawler technology through the traffic card big data platform.The data obtained through the platform better reflects the real situation of passengers taking bus lines.Secondly,the eXtreme Gradient Boosting Model is a relatively new integrated learning algorithm that has not been applied to passenger bus route prediction.The model is improved on the gradient decision tree model: it supports parallel computing,introduces first derivative and second derivative when dealing with loss function,and introduces regularization term in objective function,which has fast training speed and high prediction accuracy.The advantage of strong generalization ability.Finally,in the logistic regression model,the gradient elevation decision tree model and the eXtreme Gradient Boosting Model for the passenger bus route simulation test,the experimental results show that the eXtreme Gradient Boosting Model has faster training speed and higher prediction accuracy..In summary,this paper considers the factors such as time and date,weather information and crowds.The eXtreme Gradient Boosting Model can predict the passengers’ choice of bus routes in a complex real environment,which has positive significance for alleviating traffic congestion.