Short Term Traffic Flow Analysis And Prediction

Currently, short-time traffic flow forecasting is one of the important fields of traffic control and guidance system, and which has high accuracy is of great significance for traffic benign navigation, improving the efficiency of urban road and alleviating traffic congestion. Thus, one must deal with the short-time traffic flow forecasting in order to ensure the implementation of traffic control and guidance system. This paper analyzes the characteristics of short time traffic flow, and then uses its similarity and multi-scale characteristics to predict short time traffic flow. The main work and innovation of this paper are as follows:(1) The similarity observed at a single point on the California expressway is examined, and the similarity on the same day for adjacent weeks is higher than that on adjacent days. And the wavelet neural network (WNN), the back propagation neural network (BPN) and the least square support vector machines (LSSVM) are established using traffic flow of the same day for adjacent weeks and adjacent days as sample data respectively. After comparing the forecasting results, one can find that the prediction accuracy of using traffic flow of the same day for adjacent weeks as sample data is higher than using traffic flow of adjacent days for the three models.(2) The hybrid model of short-term traffic flow prediction based on ensemble empirical mode decomposition and wavelet neural networks is proposed. First, the traffic flow is decomposed into several sequences by ensemble empirical mode decomposition to obtain the multi-scale component. Based on this fact, and then an improved wavelet neural networks model of each component is established. Finally, the results of each component forecasting are superimposed to obtain the final forecasting result. The results show that the model has higher prediction precision, significantly better than the EMD and BPN model.(3) The hybrid model of short-term traffic flow prediction based on EEMD-Approximate entropy and WNN is proposed. After the traffic flow decomposed into multi-scale component by ensemble empirical mode decomposition, the complexity of the sub-sequences are calculated by approximate entropy to restructure and obtain new sub-sequences. And then an improved wavelet neural networks model of each component is established. Finally, all component prediction are superimposed to obtain the final result, which indicates that such model has higher prediction precision on the basis of reducing the complexity of forecasting than EEMD-WNN, significantly better than the hybrid forecasting model based on LSSVM. Besides, the short term traffic flow prediction of major holidays using the hybrid model is given in this paper. The results indicates that the hybrid model above using the training samples scaled by year can improve the prediction accuracy of short term traffic flow of major holidays.All the above studies show that using similarity and multi-scale characteristics of traffic flow for traffic flow forecasting is not only helpful to improve the forecasting accuracy but also effective for alleviating traffic congestion of city.