The Prediction Of Bus Arrival Time And Congestion Degree Based On BP Neural Network

"Intelligent traffic" is a very important part of "Smart city".It is an important research and application direction of intelligent traffic to strengthen the forecasting of urban traffic trend and release traffic forecast information timely.The accurate prediction of bus arrival time and crowded degree is an important part of the overall forecast.This paper mainly uses the BP neural network model to predict bus arrival time and crowded degree.(1)the bus arrival time factors are divided into two parts:micro and macro.The micro part regards the predicted bus as the main body of the observation.There are three types of impact of factors in micro part:the normal driving process factors,site docking factors and the traffic signal factors.The macro part regards the traffic environment as a whole system,considering its traffic status factors.In addition,this paper uses the intermediate factors as the input variables of the BP neural network model,so that it can fully cover all the influencing factors.There are three intermediate variables(ie,model input)which are chosen:1)Bus arrival time of all previous stops;2)Bus dewell time of all previous stops;3)Bus delay time of all previous stops.The choosing of these three types of features as the input variables has following advantages:1)This can integrates the factors that affect bus arrival time effectively and can cover all the impact.2)Comparing to the data(such as weather,geographic location,etc.)which are difficult to capture.The extraction and preprocessing of the input data is easier.3)The complexity of the model is reduced,and the actual operability of the system is increased.(2)based on the real GPS data of Beijing No.300 bus,combined with BP neural network model,support vector machine model,KNN and logistic regression model,the two diffrent input vectors are experimentally analyzed and compared with different measurement standards.Finally,The results of the comprehensive impact variables are more accurate,and the BP neural network is the most accurate and the calculation time is the shortest.(3)on the basis of past congestion perception mechanism,this paper analyzes the causes of internal congestion in public transportation by using the ecological perspective theory.Then a social survey is carried on.Unlike the previous text survey,this paper uses the visual perception of the crowded image survey.With the use of statistical methods,the degree of congestion in the bus is given.The actual data are analyzed experimentally,and it is confirmed that the number of bus(ie passenger flow)and bus arrival time has the same rule.(4)this paper creatively uses the same BP neural network to predict bus predict bus arrival time and crowded degree at the same time.This method makes full use of the strong nonlinearity of BP neural network.This can also reduce the complexity of the system and improve the speed of operation.The system will design a BP model for each stop of a specifically line bus.The inputs are the arrival time of all previous stops,dewell time of all previous stops and the delay time of all previous stops.And the outputs are bus arrival time and crowded degree.The number of input and output vectors for each model is also different.The results are displayed.On one hand,it can provide support for real-time traffic management and scheduling.On the other hand,information can be provided to passengers via e-stop or APP.