Research And Implementation Of Temporal And Spatial Passenger Flow Model

With the development of urban traffic system,number of passengers has a sharp increase.Travel information of these passengers is of great value.So this paper use time series model,Temporal and Spatial Model to investigate and predict passenger flow of multistation.Passenger flow has strong relation with time such as changes dynamically with time and it has periodicity.This paper use ARIMA model to fit and predict multistation passenger flow.But order is an important part of ARIMA model which can improve ARIMA's forecasting ability.So this paper uses a set of statistical methods to estimate and test ARIMA's orders.And finally compare the result of ARIMA with SVR and improve this method can generate a more accurate model.One of weaknesses of ARIMA is that it needs complex experiments to estimate parameters,which is hard to predict hundreds or thousands of stations'passenger flow.So this paper proposes Temporal and Spatial Model to solve this problem.This model can use temporal and spatial characteristic of passenger flow and predict accurately for passenger flow of multistation.Firstly,this paper propose cluster method based on investigating time property(such as trend and numerical values)of passenger flow.Experiment shows that this method can improve accuracy for predict multistation passenger flow.Secondly,the spatial property of passenger flow is analyzed in detail.A passenger flow clustering algorithm based on spatial factors is put forward based on the time factor clustering results.The second cluster algorithm is clustered from the perspective of the round-trip.The experiment proves that this method can further improve the model's ability of multistation passenger flow prediction.Meanwhile,the two clustering algorithms are implemented based on the parallel computing framework,which can effectively improve the computation efficiency and make the model fit the training data quickly.The first,second chapter focuses on the background and the significance of the research.The third chapter focuses on the characteristics of data and the key technologies of time series model and Temporal and Spatial model.Combined with the fourth,fifth chapter,this paper expounds the modeling process of two passenger flow models.The sixth chapter focuses on the implementation of the spatio-temporal model,and tests and analyzes the Temporal and Spatial Model.