Research On TAR Model Based On Bayesian Stochastic Search

The healthy development of the transport sector is one prerequisite for national prosperity.In-depth study of urban short-term traffic flow is conducive to understanding the specific cities of China's traffic laws,contribute to the"wisdom of the city"process.It is the current concern of many scholars to construct reasonable mathematical model and forecast the time series of urban road traffic flow.It is also the main focus of this paper.At the present stage,the construction and forecasting of short-term traffic flow models only consider the status of short-term traffic flow under normal traffic conditions.The model structure can not explain the non-linearity and complexity of time series.In this paper,we use the nonlinear model to study its intrinsic law,and use the threshold autoregressive model based on Bayesian stochastic search to construct a reasonable model of urban short-term traffic flow.The threshold autoregressive model is focused on the selection of threshold autoregressive models and the estimation of variables.Based on the Bayesian stochastic search model,the Bayesian stochastic search model is used to solve the problem of model selection and variable estimation.In the TAR model,the model selection problem and the variable estimation problem are transformed into the stochastic process Analysis of distribution.The posterior distribution of the random process contains information about the number of mechanisms and the threshold parameters.And combined with Gibbs and MH sampling algorithm to realize the joint sampling from model space?posterior distribution of random process?and indefinite variable space using MCMC method.In the study of the variable point model based on Bayesian stochastic search,the posterior distribution of random process is sampled with three consecutive and different distributions.Then,the TAR?2;2,2?model is modeled by using the above method to simulate the data of the natural population growth rate in our country.The results show that the conclusion of this model is in line with the present situation and statistical results of population development in C hina.It also shows that the method can well estimate the number of mechanisms,delay parameters,thresholds and the corresponding regression coefficients under the mechanism.Therefore,the TAR model based on Bayesian stochastic search is applied to the urban road short-term traffic flow.By analyzing the data characteristics and combining the road traffic situation,the TAR model with three mechanisms is constructed.The model conforms to the short time Dynamic changes in traffic flow?free flow,frequent congestion,and occasional congestion?.And the TAR?3;7,7,7?model is constructed by analyzing the short-term traffic flow of a road in Tianhe District of Guangzhou C ity,and the threshold value r₁=60,r₂=100 is obtained.The characteristics of the short-term traffic flow The At the same time,the TAR model with three mechanisms and the conventional ARIMA model are used to predict the urban road traffic.It is found that the TAR?3;7,7,7?model with three mechanisms has higher prediction accuracy than the ARIMA?8,0,0?model.In summary,the threshold autoregressive model based on Bayesian stochastic search is a reasonable way to explain the behavior of short-term traffic flow in urban roads in China.