運用 FANNC 與 BPN 評估共同基金績效

Abstract

本論文主要是比較FANNC與BPN兩種類神經網路模型運用於共同基金績效評估時的表現。FANNC為一新提出的類神經網路模型，它結合了ART與Field theory模型的特性。在我們評估共同基金績效的實證研究中，FANNC運算的速度較BPN迅速得多，RMS亦以FANNC的結果較為優異。無論於共同基金績效分類問題或績效預測問題，其結果都以FANNC較為優異。

The purpose of this paper is to compare two different approaches, FANNC and BPN, in mutual fund performance evaluation. FANNC is a newly developed neural network which combines the features in ART and field theory. In our experiment, mutual fund performance can be evaluated much faster in FANNC approach than that in BPN approach. RMS is also superior for FANNC. These results hold for both classification problems and for prediction problems.