台灣股價指數預測模型之探討

An investigation of the forecasting models for stock price variation in Taiwan

楊雯寧(Wen-Ning Yang)

指導教授 ： 張百棧

元智大學/工程學院/工業工程與管理學系/碩士(2002年)

摘要

在過去所探討股價趨勢之研究中，主要分為兩大派別：一以技術面的觀點，藉由股價過去的走向以預測未來之波動；二以基本面的角度，分析各種影響股價之總體經濟因素以預測股價。本研究分別以基本面之經濟因素與技術面之各項指標為依據，以台灣股票市場發行量加權股價指數為分析對象，利用逐步迴歸法篩選與股價指數變動有顯著關係之變數，再依此建立多元迴歸分析、倒傳遞類神經網路、以及自我迴歸移動平均整合模式（ARIMA），將三種預測模型所得到之結果作為模式合併之根據，並透過一動態權重值α結合分別以基本指標與技術指標為考量所得到的月股價指數與日股價指數之預測值，以建立混合式預測模型，以期更準確的預測每日股價指數。結果發現混合式模型平均誤差為0.493%，相較於其他三種預測模型中預測能力最佳之多元迴歸分析的平均預測誤差0.703%，其改善效益達29. 875%，因此混合式模型為本研究案例對於預測未來股價波動適用性最佳之預測模型。

關鍵字

多元迴歸分析 ； 倒傳遞類神經網路 ； 自我迴歸移動平均整合模式 ； 基本指標 ； 技術指標

並列摘要

The investigation to the stock market index in the literature could be divided into two aspects. The first one is the technical analysis by using the historical trend of the stock market index to predict the future fluctuation. The other one is the basic analysis; it analyzes the factors that affect the macroeconomics to forecast the stock market index. In this research, the technical analysis and the basic analysis were applied to investigate the trend of the weighted stock market index in Taiwan. Stepwise regression analysis was first used to identify the key variables that affect the trend of the stock market index significantly. According to the identified variables, three models, i.e. a multiple regression analysis model, a backpropagation neural network, and an autoregressive integrated moving average model were built. A hybrid model that integrates the technical and basic analyses was developed and expected to forecast the stock market index more accurately. The result has showed that the average error of the hybrid model is 0.493%. The best performance of the other three models is 0.703%. The experimental results shows that the improvement in the hybrid model attains 29.875%. The result indicates that the hybrid model is the best one in forecasting the future fluctuation of the stock market index in this research.

並列關鍵字

Multiple regression analysis ； Backpropagation neural networks ； Autoregressive integrated moving average models ； Basic index ； Technique index

參考文獻

[23] 楊志祥，「類神經網路與多元迴歸之比較研究：以台灣股票行為之預測為例」，私立元智大學，工業工程研究所碩士論文，民國86年。

[26] 廖廣義，「以類神經網路預測股價指數漲跌」，私立元智大學，工業工程研究所碩士論文，民國88年。

[1] Baba, N. and M. Kozaki, “An Intelligent Forecasting System of Stock Price Using Neural Networks,” IJCNN-92, Vol. 1, pp. 371-377, 1992.

[2] Box, G. E. P. and Jenkins, G. M., Time Series Analysis Forecasting and Control, Holden-Day, San Francisco, 1976.

[3] Chi, S. C., Chen, H. P., and C. H. Cheng, “A Forecasting Approach for Stock Index Future Using Grey Theory and Neural Networks,” IEEE International Joint Conference on Neural Networks, pp. 3850-3855, 1999.