Research On Short-Term Traffic Flow Prediction Based On Shuffled Flog Leaping Algorithm And Wavelet Neural Network

With the expansion of China's urbanization construction,traffic congestion is becoming more frequent,which greatly increases people's travel time on the road.Intelligent transportation system can effectively alleviate traffic congestion through real-time road condition feedback,as the important basis of ITS,the study on fast and accurate short-term traffic flow prediction has deep practical significance.Considering the uncertainty and nonlinear characteristics of short-term traffic flow data,the prediction effects of many models are not good.Wavelet neural network(WNN)combines the advantages of wavelet transform and neural network,has good characteristics in time and frequency domain.Shuffled flog leaping algorithm(SFLA)is a recently proposed intelligent group algorithm,which shows good robustness and global convergence,and can find the optimal solution quickly.Therefore,this paper combines the shuffled flog leaping algorithm with the wavelet neural network,use the shuffled flog leaping algorithm to optimize the weight of the network,and use the improved shuffled flog leaping algorithm to improve the speed and accuracy of prediction.Finally,the radial base function(RBF)neural network is added to further improve the prediction effect of short-time traffic flow signal.The main work of this paper is:(1)The traffic flow data of each data set are pretreated,including repairing abnormal data,wavelet de-noising based on improved layer-by-layer adaptive threshold and exponential change threshold function,phase space reconstruction and normalization processing,and then the preprocessed data set is predicted by using the WNN model.Experimental results show that the model can simulate the trend of short-time traffic flow,but the prediction effect still needs to be improved.(2)Combine the shuffled flog leaping algorithm with the wavelet neural network,the SFLA-WNN model is constructed,and replace the initial weight of the WNN network with the optimal solution obtained by SFLA optimization to avoid the defect that the network falls into the local minimum.The experimental results show that the combined model can effectively improve the prediction accuracy of the network.On this basis,the shuffled flog leaping algorithm is improved to optimize the individual moving step size.Experiments show that the ISFLA-WNN model can significantly reduce the running time under the condition of ensuring the accuracy and stability of short-term traffic flow prediction.(3)Add the radial basis function neural network,the ISFLA-WNN-RBF model is constructed.Divide the traffic flow data set into linear and nonlinear parts,and use the SFLA-WNN and SFLA-RBF models to predict each part separately,superimpose to get the final result.Experiments show that short-term traffic flow prediction based on ISFLA-WNN-RBF model can effectively reduce errors,improve prediction accuracy,and has excellent stability and prediction effects.