A Novel Design of Hybrid Polynomial Spline Estimation and GMDH Networks for Modeling and Prediction

Abstract

GMDH algorithm can well describe the internal structure of objects. In the process of modeling, automatic screening of model structure and variables ensure the convergence rate.This paper studied a novel design of hybrid polynomial spline estimation and GMDH. The polynomial spline function was used to instead of the transfer function of GMDH to characterize the relationship between the input variables and output variables. It has proved that the algorithm has the optimal convergence rate under some conditions. The empirical results show that the algorithm can well forecast tax revenue.