Research on Freeway Investment Risk Assessment Model Based on Variable-Structure Neural Network

Abstract

Neural network models have widely been applied in assessment and perdition of economic and social fields, including risk assessment. Thus, it becomes a subject for the theory of neural network to study how to improve accuracy in the premise of ensuring convergence rate of BP (Back Propagation) neural network. On the basis of recent studies and disadvantages of traditional BP neural network, in terms of structural optimization to improve accuracy, the paper presents a variable-structure neural network where it is re-linking randomly process from neurons of input layer to neurons of output layer and from neurons of hidden layer to neurons of output layer. Secondly, the variable structure neural network of re-linking random process is applied in freeway investment risk assessment. Results of a cast indicate that the proposed model is sufficiently reasonably.