Research On Intelligent Bus Passenger Flow Inferring Method Based On IC Card Data

Urban public transportation is an important part of urban public transportation system.Developing urban transportation can improve various transportation problems brought about by urban development.The passenger flow between bus stops reflects the time and space distribution characteristics of passengers.It is an important basic data for bus system evaluation,bus scheduling and bus network optimization.The bus IC data contains a large number of passengers' travel data,which is the main data resource for obtaining bus passenger flow.However,in most cities,IC data only records passengers' information on boarding,and lacks relevant data for getting off the bus,resulting in the inability to directly obtain passenger flow How to accurately infer the passenger's alighting stops and obtain the passenger flow between bus stations is an urgent problem that needs to be solved,which is of great significance to the optimization of urban bus system.This paper proposes a two-stage bus passenger flow inference method based on bus IC data.Firstly,two strong rule inference methods are designed for two common modes of travel:transfer inference and round-trip inference.By introducing time and space constraints and rules,eligible passengers were found and the records that met the requirements were inferred with high accuracy.Then,the problem is regarded as the sequence tagging problem,the basic travel characteristics of the passengers are extracted,and the training set and test set are constructed based on the results obtained in the first stage.This paper uses the Recurrent Neural Networks nesting Condition random field model,the trained model can predict all records' alighting stops.At the end of the method,in order to achieve optimal results,the results of the two stages are integrated.In this paper,experiments were carried out on three bus lines,and millions of real-world bus IC data were used to verify the effectiveness of each step of the method,and compared with various methods in multiple angles.The experimental results show that the two rule inference methods proposed in this paper have high accuracy and can cover a certain proportion of records,which makes it possible to use the sequence tagging method for low-cost bus passenger flow inference.The framework of the combination of rule and sequence tagging algorithm proposed in this paper achieves high precision in different lines,different time periods and different types of passengers,which are far better than the traditional trip chain model.The Bi-LSTM-CRF used in this paper performs best.The experiment proves that the proposed method has the characteristics of low cost,accuracy and robustness.