建構二階段模型預測台灣加權股價指數與台灣50指數

The Implementation of Two-stage Prediction Model for Taiwan Weighted Stock Index and Taiwan Top 50 Exchange Tracker Fund

詹佩茹(Pei-ru Chan)

指導教授 ： 周宗南

朝陽科技大學/管理學院/財務金融系/碩士(2009年)

摘要

欲預測股價的漲跌幅有一定困難度，原因為影響股市漲跌因素諸多，有總體之經濟因素；亦有個體之經濟因素，均會影響股價表現。自2008年美國華爾街的金融危機引發全球金融風暴，全球經濟一瞬間跌落谷底，股市一蹶不振，公司遭受景氣不佳的影響下，都會選擇裁員或減薪，來精簡人事費用的預算，造成失業率不斷高升，「無薪假」風潮席捲大小規模之公司，即使是高科技大廠亦難以倖免，員工無薪休假所引發的勞資爭議更是層出不窮。
雖然全球股市持續大跌，卻正是小額投資人逢低進場的時機，不過單單掌握基本分析與技術分析還是不夠，有鑑於此，本研究使用人工智慧的方式，預測股市的走向，主要以台灣加權股價指數與台灣50指數為研究標的，加入基因演算法、類神經網路、基因規劃法、灰色預測與灰色決策所建構出的模型，經過短期、中期與長期的訓練與預測，研究發現，灰色預測以級比調整過後之預測結果較佳，其次為基因規劃法。

關鍵字

基因規劃法 ； 灰色理論 ； 類神經網路

並列摘要

To estimate and predict the fluctuation of a stock market is difficult because lot of variables including both macroeconomic and microeconomic factors influence the stock price. Due to the world financial storm resulted from Wall Street financial crisis, global economy plunged to the valley bottom in a flash and the stock market collapsed and met with many a setback. Many companies suffer the unprosperous influence and therefore choose the ways to layoff staffs, to cut salaries and even shutdown branches to reduce the operation cost for the reason of budget. As a result, the rate of unemployment increases and unpaid leave sweeps across large and small scale companies, even a large High-Tech company strives to keep alive.

As the global stock market slumps continuously, it is exactly a good opportunity of marching into the arena at a low price for investors. However, to acquire maximum returns, the investors need not only basic analysis and technological analysis but also other effective approaches. This research work applies artificial intelligence methods to predict the trend of the Taiwan Weighted Stock Index and Taiwan Top 50 Exchange Tracker Fund. Various approaches including Genetic Algorithm, Artificial Neural Networks, Genetic Programming, Grey Prediction with Class Ratio and Grey Decision are employed in this study. The results suggest that the Grey Prediction with Class Ratio performs better than others and the Genetic Programming ranks as second through the periods of short-time, middle-term and long-term training and prediction.

並列關鍵字

Grey Systems ； Aritificial Neural Networks ； Genetic Programming

參考文獻

25. 陳美淑(2004)。以基因規劃演算法預測公司財務危機。明新科技大學企業管理系，新竹。

6. 李佩芳(2005)。整合遺傳演算法與模糊理論以建構財富管理決策模型之研究。朝陽科技大學財務金融系研究所碩士論文，台中。

7. 李惠妍、吳宗正與溫敏傑(2006)。迴歸模式與類神經網路在台股指數期貨預測之研究。經營管理論叢。2，83-99。

15. 姚凱齡(2008)。應用限制滿足式遺傳演算法於股票投資策略制定。大同大學，資訊經營系碩士論文，台北。

1. Allen, F. & Karjalainen, R. (1999). Using Genetic Algorithms to Find Technical Trading Rules. Journal of Financial Economics, 51, 245-271.