RBF Neural Networks Optimization Algorithm and Application on Tax Forecasting

Abstract

Accurate tax revenue forecasting has become a most important management goal, however, tax revenue often presents nonlinear data patterns. Therefore, a rigid forecasting approach with strong general nonlinear mapping capabilities is essential. The genetic algorithm has been used to select the parameters automatically for support vector machine. Then a RBF neural network has been built based on support vector machine and genetic algorithm, which helps to form a forecasting model system of RBF neural network optimization algorithms. As a result, this algorithm can avoid not only the shortcomings of traditional algorithm which is easy to get local minimal value, but also a large number of experiments or experiences which are needed to pre-specify network structure. 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.

DOI: http://dx.doi.org/10.11591/telkomnika.v11i7.2199