A travel time prediction method: Bayesian reasoning state-space neural network

According to the prediction model of neural network training methods to slow convergence speed, training for a long time and difficult to control the complexity of weights updating, this paper puts forward Bayesian reasoning state-space neural network, using termination conditions control its training and the confidence interval restrained by control factor standard the results. Using this method can accelerate convergence, shorten the training time and maintain stability. With traffic data by Bin He road of Shenzhen in September 2007 to verify this model, the experiments show that this model can shorten the time of training, and has good robustness and accuracy.