Investigation of Favorable Neural Network Methods to Estimate Traffic Components

Neural networks provide the opportunity to estimate specific components of engineering problems. They are decomposed complex problems into different parts. Thus, it can be easy compete with each them through neural networks. In this paper, was purposed average speed a 6-line road’s cross-section by observed traffic variables, such as numbers vehicles and occupancy values, using radial basis function network (RBFNN), generalized regression (GRNN) feed-forward back propagation (FFBPNN) models. A comparison fulfilled between checked against multivariate linear (MVLR), conventional statistical model. After simulation networks, results show that forecasts were obtained under same conditions. The best forecasting is made FFBPNN, GRNN, RBFNN, respectively. When compared FFBPNN performs better than MVLR, but GRNN RBFNN perform lower it.