Method On Passenger Flow Forecasting Of Bus Station Based On Location And Boarding Counts

Under the background of increasingly prominent problems such as traffic congestion,traffic pollution and traffic accidents,public transport priority development has been one of the most effective means to alleviate urban traffc problems.Development of public transportation also can promote the development of economy.It has become the main body of urban passenger transport system and one basic industry of national constructions strongly supported.Traditional statistic patterns of bus passenger flow have obvious shortcomings such as corrplex investigation procedures,high cost and low accuracy.The data collected by Automatic Fare Collection System of conventional bus for data mining is analyzed and the travel characteristics and laws of bus passenger flow to predict bus passenger flow are researched more efficiently and more accurately in this paper.The existing traffic information resources are used reasonably.On the basis of related research literatures,three aspects of research results at home and abroad were reviewed in this paper,including estimation of passenger flow at stations,characteristics analysis and traffic short-term forecasting method.The existing research results are summarized and compared,and the research focuses of this paper are pointed out.Firstly,the characteristics of data are analyzed,original data is preprocessed by data filtration and data deletion according to the characteristics and passenger flow estimation demand.Data clustering analysis was carried out on the Automatic Fare Collection System records.Data of multi-source information is fused by time matching to match location and boarding counts information,including Automatic Fare Collection System data,GPS data,static network data and bus dispatching information.Station and AFCS record are matched,and then the passenger flow of station is estimated.Secondly,Jinan Advanced Public Transportation Systems data is taken as an example to validate the matching algorithm based on multi-source data fusion.Direction unequilibrium and balance between supply and demand are researched to analyze the route characteristics.Commuting characteristics and time unequilibrium are researched to analyze the station characteristics.According the station characteristics,the distinctive stations of routes are selected to prepare for the passenger flow prediction.Finally,variables analysis of the factors effecting passenger flow at stations is carried out,time series model and the modified BP neural network model are selected to estimate the passenger flow of distinctive stations of routes.Prediction accuracies of different prediction models on different kinds of stations are compared,then station traffic demand forecasting conclusion is concluded.