| Summary: | 碩士 === 大葉大學 === 國際企業管理學系碩士在職專班 === 96 === The goal of this study is applying Grey prediction, radial basis function neural network (RBFNN) and Grey-RBFNN to predict the price of electronic sector index fu-tures and the finance sector index futures, and to compare their prediction accuracies. Grey model is a simple approach with acceptable prediction accuracy whereas radial basis function neural network is a tedious manipulation with high prediction accuracy. The new model Grey-RBFNN combining Grey and RBFNN is proposed and tested. It utilizes the GM(1,1) prediction as one of the RBFNN inputs. Prediction performances of all three algorithms are compared by using data of three single nation equity funds, JF Taiwan Fund, JF Japan New Generation Fund, and JPM New America Trust Fund from JP Morgan Asset Management, Taiwan. The period of samples is from January 2001 to December 2005. The results reveal: (1) GM-RBFNN is better than RBFNN, and RBFNN is better than GM(1,1). (2) By adopting the multi-regression analysis to com-paring with the residual of GM(1,1)、RBFNN、GM(1,1)-RBFNN and multi-regression. The fitness results indicate the GM(1,1)-RBFNN is the best, and the GM(1,1) is the worst. (3) After training phase, the prediction accuracies of GM-RBFNN and RBFNN are all improved.
|
|---|
