Based On Firefly Algorithm And RBF Neural Network Of Highway Traffic Flow Prediction

Based on the overall trend of highway traffic state for effective control, can effectively alleviate the congestion of the highway and prevent the occurrence of traffic accidents. Highway traffic forecast is one of the key issues, and the traffic flow has the characteristics of complexity, nonlinearity, make the traditional forecasting model cannot be accurately predict the change trend of traffic.Radial basis function(RBF) neural network as a new neural network model, it has a fast learning speed, strong convergence, self- learning and adaptive, and there will be no local minimum value problem, it is the best approximation of nonlinear function. So the RBF neural network research become the hot issue in the traffic prediction model, but the parameters of RBF neural network hidden layer neurons is a difficult problem to determine. As a result, the firefly algorithm was used to optimize RBF neural network is proposed in this paper, to determine the center, width of neurons in hidden layer and weights between hidden layer neuron and output layer neuron.This article first introduces the principle and structure characteristics of BP neural network and RBF neural network, and made a contrast the two kinds of neural network model, the results show that the RBF neural network has the characteristics of simple structure, good convergence performance. Secondly introduces the commonly used two kinds of intelligent swarm optimization algorithm, genetic algorithm, particle swarm optimization algorithm, analyzed the principle and advantages and disadvantages of both. Then make a detailed analysis and explanation for the principle of firefly algorithm and the application of firefly algorithm used in the neural network. And put forward the RBF neural network predict ion model based on the firefly algorithm optimization. Finally, design three neural network model optimized genetic algorithms, particle swarm optimization and firefly algorithm. And train three neural network models with real highway traffic flow. And use the trained three models to predict traffic flow. From the prediction accuracy and training efficiency and generalization ability three aspects analyzed the results. Results show that the RBF neural network optimized by the firefly algorithm prediction model has higher prediction accuracy, faster training speed, and better generalization ability.