Improved Particle Swarm Optimization For Elman Neural Network Traffic Flow Prediction

With the prosperity of national economy and the improvement of people’s living standard,China’s automobile industry develops rapidly,and the number of automobiles also increases dramatically,which poses a huge obstacle to the development of road traffic and environmental protection.In order to alleviate the congestion caused by too many vehicles,traffic control should be implemented to realize real-time management of vehicles on the road.The intelligent transportation system(ITS),as an important way to solve traffic problems,has developed rapidly in recent years.Combined with advanced information technology and traffic theory,it can realize the effective control of road traffic and the ability of traffic induction.As a branch of intelligent traffic,short-term traffic flow prediction plays an important role in traffic forecasting.Due to the rise of neural network,various network models are applied to the prediction of road traffic state.This paper proposes an improved particle swarm optimization dynamic Elman neural network for traffic flow prediction.The main research work is as follows:(1)Elman neural network is a feedforward neural network,which adds its own underlayer to the hidden layer of network structure to make it also applicable to the dynamic model.However,it is prone to get trapped in the local minimum and the convergence speed of the algorithm is affected by the number of layers in the network.Therefore,the combination of swarm intelligence optimization algorithm and Elman neural network is introduced to rationally utilize the global search performance of the particle swarm optimization algorithm.(2)For the traditional particle swarm optimization algorithm,according to the particle velocity and position formula,the algorithm can be improved by inertia weight,learning factor and other parameters to improve the performance of the improved algorithm.In this paper,the improved particle swarm optimization Elman combination model is used to predict the traffic flow and congestion degree of a section of xi’an south second ring road.(3)By analyzing the advantages and disadvantages of the traditional particle swarm optimization algorithm,this paper proposes improvement from three aspects: linear decreasing inertia weight,removing particle velocity term,linear decreasing inertia weight and dynamic adjustment learning factor.Through experiments,the ratio of prediction resultsof the three improved network models is obtained.Among them,a particle swarm optimization algorithm with linear decreasing inertia weight and dynamic adjustment learning factor to optimize Elman neural network has more accurate prediction effect.