In order to reveal the inherent dynamic characteristics of traffic flow

the Lempel-Ziv algorithm in symbol dynamics is used to calculate the time series complexity of traffic flow at different time scales. Based on the analysis method of correlation theory

the correlation of traffic flow in time and space is explored. Aiming at the problem that the classical single-regime function model can’t accurately characterize the relationship of traffic flow parameter under different traffic condi-tions

and a two-stage traffic flow parameter relationship model is proposed.The traffic flow data collected by the Pe MS system is used to fit the relationship of traffic flow-density with the nonlinear least squares method. The experimental results show that the fitting performance of our model is better than the other four models. It can be seen from the RMSE that the standard error of the model is 7. 6 % lower than that of single-regime 3 PL model. Which indicates that the model can better represent the parameter relationship of different traffic states. The established parameter model has value of practical application for traffic state discrimination and short-term prediction.