Application in Tax Forecasting of Main Component Method

Artifical neural network is suitable for non-linear systems and is popularly applied in economic forecasting.But there are so many indexes and few historical numbers in ecoomic forecasting that weaken the generalization performance of the neural network.This article puts forward a tax forecasting model by use of main component method.Results indicate that reforecasting after the main component method can drop the scope of the neural network and improve its generalization performance.