應用分類元系統於台灣股票市場指數漲跌預測

Abstract

股市莫名的波動、非線性或近似渾沌的特質長久以來一直是股市投資者及財務研究學者最有興趣的課題。故如何活用電腦科技、財工數學和計量經濟等要素，以產生最適宜的財務投資決策，來降低人為錯誤，並擴大獲利機率為本研究所積極思考之方向。 在資訊科技的應用上本研究強調採用1、人工智慧『分類元系統(eXtended Classifier System,XCS)』之動態知識學習，因該方法學是一個以規則為基準的主動式機械學習（machine learning）系統，並具適應動態環境學習的特性，故可以更貼近人類學習、決策方式及瞭解股票脈動。2、運用『群組決策』機制，結合『知識分享』的概念，利用雞尾酒療法提升決策準確率，是故本研究運用此兩大特色自行設計介面功能完整之智慧型金融投資決策系統。在系統效能評估上，本研究以單一技術指標、單一分類元系統、Buy & Hold，銀行6年期定存為對照組，實證指出系統在預測的準確率、投資累積報酬率、投資報酬波動等模擬績效上均較對照組有顯著的效果。

The fluctuation, nonlinear and chaotic characteristics of the stock market have long been an interesting topic for investors and financial researchers. This study attempts to take the advantages of computer technology, financial mathematics, and econometrics to make reasonable investment decisions, reduce man-made errors or mistake and increase the opportunity to make greater profits. This Extended Classifier System (XCS) in this study is implemented, which has new features such as dynamic learning and group decision making. An empirical study is done by comparing the profitability of the proposed system with that of investment strategies made by using single technical index, XCS with single classifier, buy and hold, and interest revenue at 6-year bank lending rate. The proposed system shows superior performance in percentage of accuracy, rate of accumulated return, and variance of return.