Research On Short-term Traffic Flow Forecasting Based On Chaos Characteristic Analysis And MEA-LSTM Model

An efficient intelligent traffic control depending on Accurate and efficient short-term traffic flow forecasting that can also promote the development of intelligent traffic and a key measure to accelerate the process of intelligent traffic.Therefore,how to accurately analyze the characteristics of traffic flow,perform accurate data repair on missing data,establish an accurate traffic flow prediction model,and provide an effective data foundation for smart transportation,and can greatly promote the development of the transportation industry..Firstly,in this paper,the traffic flow data screening process and noise reduction method are studied.After in-depth analysis of the PeMS database,the representative traffic flow data was selected,and the mechanism of data abnormalities and missing was analyzed,and the NNSADE method was proposed to repair the data,the other two models are compared show that the NN-SADE model has the highest accuracy.After obtaining high-quality data,the maximum Lyapunov exponent is used to determine the chaotic characteristics.Then,by using phase space reconstruction technology and Bayesian estimation theory,all the features of the threeparameter sequence are fused in the high-dimensional space,and the effective data with better performance is obtained.Secondly,the prediction model based on time series method is studied.Three times exponential smoothing and weighted moving average method and ARIMA were selected to forecast traffic flow,and then chooses three error evaluation indicators to compare the forecast accuracy and so on.It shows that ARIMA has better accuracy in the theoretically driven model.Thirdly,Research on predictive models based on data-driven model.This article chooses a deep learning method suitable for processing time series: LSTM model.After analyzing and predicting the method,the MEA algorithm can be used to divide the data into several subgroups for optimal search,so MEA-LSTM model is proposed.And LSTM and MEA-LSTM are used to analyze the data.The results show that the MEA-LSTM model has better accuracy in the data-driven model,it also performs better in terms of predicting speed.Finally,the two kinds of different methods are compared.In this paper,the MEA-LSTM model is compared with the ARIMA model which has the best performance in time series method.The results show that the MEA-LSTM model is better in terms of prediction accuracy,and the MEA-LSTM model is suitable for a variety of data and has a higher generalization ability,while the ARIMA model still has greater advantages in terms of computational efficiency.