應用分群技術與基因演算法建構以趨勢為基礎之股市投資策略

Using Clustering and Genetic Algorithms to Construct an Investment Decision Model Based on Stock Trend

吳忠敏(Chung-Min Wu) ； 周聖鈞(Sheng-Chun Chou)

《臺北科技大學學報》 45卷2期 (2013/05) Pp. 49-67

摘要

本研究提出一個新的股價趨勢分群與趨勢投資決策模型，利用基因演算法的最佳化能力，建構出隨股價趨勢波動之投資策略，以提供投資人操盤時的參考。股價趨勢分群模型運用分群技術將股價分為漲勢、盤整、跌勢等三種趨勢，克服過去僅能以股價或技術指標圖形訊息判斷趨勢造成過於主觀且無法量化的問題。本研究以台灣加權股價指數（TAIEX）為研究對象，探討趨勢策略於股市投資決策之績效，並與市場上常用的投資策略做比較，以驗證趨勢策略之有效性。實證結果顯示，股價趨勢分群模型能夠正確的將股價分為三種趨勢；另外，基因演算法所建構之趨勢策略模型於多頭或空頭期間，投資績效皆優於葛蘭碧法則策略、KD指標策略與買入持有策略，是一個穩定且有效的投資策略。

關鍵字

分群技術 ； 基因演算法 ； 股價趨勢 ； 投資策略

並列摘要

Most of investors use similar investment approaches in the stock market, no matter in bull period or in bear period. It may lead to unstable performance and relatively low profit. In this research, new stock trend clustering model and investment decision model have been proposed. By adopting genetic algorithms, the new investment decision model would recommend on strategy of investment. The new stock trend clustering model identifies the trend of stock price in terms of ＂up trend＂, ＂flat trend＂ and ＂down trend＂, in order to overcome limitations when diagnosing trend by referring to stock price, technical indicators and graphic messages only. This paper is based on daily stock trading records from the Taiwan stock exchange capitalization weighted stock index(TAIEX), which includes stock prices and trading volume from January 1, 1991 to March 30, 2012. The paper has reviewed the performance of investment strategy based on the new stock trend clustering model. Also, common investment strategies have been compared with the new model, in order to prove its effectiveness. This paper has proved that the new stock trend clustering model could identify stock price in three trends precisely. On the other hand, the investment decision model based on genetic algorithm has a better performance than ＂Granville rules＂, ＂KD indicators rules＂ and ＂buy and hold strategy＂. It has demonstrated the new models are reliable and effective investment strategy.

並列關鍵字

Clustering ； Genetic Algorithms ； Stock Trend ； Investment Strategy