Traffic Prediction Method Based on RBF Neural Network with Improved Artificial Bee Colony Algorithm

Abstract

In order to increase the precision of traffic flow prediction, a short-time prediction method based on RBF neural network is put forward and an improved artificial bee colony (ABC) algorithm is used to train the RBF neural network. The improved artificial bee colony algorithm is described as follows: firstly, the order of employed bees is arranged according to the fitness value in artificial bee colony algorithm. Only half of the employed bees having higher fitness value are kept and the rest convert to the scout bees. Secondly, the search strategy of onlooker bees is adjusted by the order of employed bees so as to generate the candidate food source around the optimum solution in population. Thirdly, to the scout bees converted by the employed bees, they generate uniformly distributed food source in given range. The improved ABC algorithm is applied to train the center and the width of Gaussian function and the weight between the hidden layer and the outer layer of RBF neural network. To validate the algorithm, we compare the algorithm presented in this paper with PSO RBFNN model. The simulation results show that the prediction error of the improved ABC RBFNN model is smaller than that of PSO RBFNN and the probability of being trapped into local minimum is largely reduced.