The short-term traffic volume forecasting for urban interchange based on RBF Artificial Neural Networks

According to the field data, analyzing the difference between the traffic flow characteristics of the interchange and that of the road base section. Researching the influence of the total traffic volume in weaving segment, the weaving traffic volume ratio and the weight vehicle percentage of the lane N2 on the traffic volume in the merging area of the interchange, the results show that the traffic volume in merging area has nonlinear relationship with the three factors above. Then using the quality that Artificial Neural Network has the characteristics of nonlinear mapping, dealing with parallel data and self-studying ability to research the forecasting method of traffic volume in interchange merging area, and designing a RBF Artificial Neural Networks with three input nerve cells and one output nerve cell. Based on the field data to train the network, and to verify the RBF Artificial Neural Networks based on another group of field data by comparing the simulation data with the field data, the results verified show that the method, using RBF to forecast the traffic volume in merging area of the interchange, is feasible, and the accuracy is rather high. The conclusion of this paper is useful for the control and management of the interchange.