| With the rapid development of China’s economy,a sharp increase in car ownership has taken place.However,too many vehicles also increase the pressure of the existing transportation system,which can lead to traffic congestion,traffic accidents and environmental pollution.In this case,the intelligent transportation system emerges as the times require,and its key technology is to establish an accurate short-term traffic flow prediction model to predict the traffic flow in the future period and thus to judge ahead of time the possible traffic situation in the future,which can facilitate the transportation department to adjust in time and improve the road capacity.It is of great significance to improve the traffic environment with this system.Based on the analysis of the existing traffic flow prediction models at home and abroad,summarizing the advantages and disadvantages of each prediction method,according to the characteristics of traffic flow itself,and combined with the theory of decomposition and reconstruction,a prediction model based on complementary ensemble empirical mode decomposition and genetic least squares support vector machine is proposed in this paper to achieve more accurate prediction,and verify its effectiveness.The work done is as follows:(1)This part introduces the data source of traffic flow used in the paper,classifies the types of data errors,takes targeted measures to repair the error data,and uses wavelet transform to de noise the data,so as to obtain the time series of traffic flow with relatively small errors and more realistic situation,which can provide reliable data support for the prediction model and facilitate the following research and prediction.(2)This part describes the basic theory of traffic flow.In order to better study the internal movement trend of the time series of traffic flow,it reconstructs the phase space,expands the one-dimensional data to high-dimensional space,then selects the C-C algorithm to calculate the parameters needed for reconstruction and calculates the Lyapunov exponent of the time series of traffic flow in the experiment.The result is greater than 0,which indicates that the selected experimental traffic flow is chaotic and suitable for short term forecast.(3)To predict the experimental data,the GA-LSSVM established.The results show that the prediction accuracy of the optimized LSSVM is improved,but the ability to process data with large volatility needs to be further improved.(4)Because of some characteristics of traffic flow,such as non-linear,non-stationary and volatile,the decomposition method is thus used to decompose the traffic flow series to reduce the impact of the fluctuation of the original data on the prediction.After comparing the reconstruction errors,it is prove that the time series of traffic flow after the complete ensemble empirical mode decomposition is more stable and suitable for prediction measurement.(5)The internal dynamic characteristics of the decomposed component can be restored by reconstructing the phase space on the basis of the established traffic flow prediction model based on complete ensemble empirical mode decomposition used to decompose the time series of traffic flow.The reconstructed components are inputted to the least squares support vector machine optimized by genetic algorithm to construct the traffic flow prediction model based on the complementary integrated empirical mode decomposition and the least squares support vector machine.According to the control experiment,compared with the traditional single prediction model,the proposed traffic flow prediction model based on the complete ensemble empirical mode decomposition and the genetic least squares support vector machine has the MAE of 13.46,the MAPE of 3.57% and the MSE of 282.23,which are lower than the control model,It can meet the demand of short-term traffic flow prediction,which shows that the combined model can improve the prediction accuracy. |
PDF Full Text Request