Short Term Traffic Flow Forecasting Based on Improved Echo State Network

The echo state network reserve pool is a random connection between the neurons, which makes the strong coupling between the neurons limit the richness of neuron dynamics, impacting prediction accuracy. In view of the above problems, a new echo state network with the characteristics of world small is proposed. Small world topology is generated by using a new algorithm based on neuron space growth, then the nodes in the network are sorted in a new way, finally, the connection between the physical nodes in the network and their interaction is mapped to the inner neurons of the new echo state network reserve pool. Simulation experiments show that, the dynamic characteristics of the improved echo state network are more abundant than the original ESN, and the accuracy of the prediction is better than that of the conventional ESN.