RBF Neural Networks Optimization Algorithm and Application on Tax Forecasting
چکیده
Tax plays a significant role in China's rapid economic growth. Therefore it is of particular importance to improve the predictability and accuracy of the tax plan. The tax data is characterized by being so highly nonlinear and coupling that it is difficult to be represented by using an analytical mathematical model in an accurate way.In this paper, a new optimization algorithm based on support vector machine and genetic algorithm for RBF neural network is presented. First the genetic algorithm is used to select the parameters automatically of support vector machine, and then support vector machine is used to help constructing the RBF neural network. The network on basis of this algorithm can be applied to nonlinear system identification like tax revenue forecasting. Case study on Chinese tax revenue during the last 30 years demonstrates that the network based on this algorithm is much more accurate than other prediction methods.