橢圓形模糊系統於台灣股市股價預測之應用

橢圓形模糊系統(ellipsoidal fuzzy system,EFS)為一可加模糊系統(additive fuzzysystem)，具有非監督式(unsupervised)學習與監督式(supervised)學習之功能，此模式已成功的應用於控制系統與型態辨識(pattern recognition)問題之研究。本研究将以橢圓形模糊系統為基礎，應用可加模糊系統對不確定問題之趨近(approximating)能力，並以非監督式群聚(clustering)資料與監督式調整(tune)資料的特性，利用共軛梯度(scaleconjugate gradient)之監督式學習法則，以預測台灣股票市場之股價。本文所提出之橢圓形模糊系統将應用於預測台灣七種不同類股之七家上市公司股價，結果顯示橢圓形模糊系统於預測短期股價表現較其他三種現有之方法好。

關鍵字

橢圓形模糊系統；可加模糊系統；非監督式學習；監督式學習；共軛梯度學習法

並列摘要

Forecasting of stock market is one of the most important topics in business.The ellipsoidalfuzzy system learning with and without supervision has been successfully applied in controlsystems and pattern recognition problems.In this study,the ellipsoidal fuzzy system ismodified to examine the feasibility for predicting stock market in Taiwan.A scaleconjugate gradient learning method is borrowed to speed the training process in supervisedlearning.Three existing forecasting approaches are used to compare the performance.Numerical results show that the ellipsoidal fuzzy system outperforms the other threemethods in forecasting stock prices in Taiwan.

並列關鍵字

ellipsoidal fuzzy system ； scale conjugate gradient ； supervised learning

參考文獻

Baba, N.,M. Kozaki(1992).An intelligent forecasting system of stock price using neural networks.Proceedings of International Joint Conference on Neural Networks.1,371-377.

Google Scholar

Callen, L.,C. Y. Kwan,C. Y. Yip,Y. Yuan(1996).Neural network forecasting of quarterly accounting earnings.International Journal of Forecasting.12(4),475-482.

Google Scholar

Cheung, Y. M.,Z. H. Lai,L. Xu(1996).Application of adaptive RPCL-CLP with trading system to foreign exchange investment.Proceedings of IEEE International Conference on Neural Networks.4,2033-2038.

Google Scholar

Cristea, A. I.,T. Okamoto(1998).Energy function construction and implementation for stock exchange prediction NNs.Proceedings of International Conference on Knowledge-Based Intelligent Electronic Systems.3,403-410.