修正濾嘴法則於台灣股市之報酬率研究

Abstract

濾嘴法則是一個有效率的股市交易法則，其績效表現在過去的研究中已被廣泛地驗證，然而實驗結果顯示，該法則所產生的交易點並非都具有良好的獲利能力。所以本研究的目的在於修改原濾嘴法則，以提高獲利能力。而修改的基本概念在於以分群法找出具有高報酬之交易點的相似特性，作為強化濾嘴法則的參考依據。故本研究利用資料探勘技術中之決策樹來解決此分群問題，找出其相似特性來發展修正式濾嘴法則。最後，本研究以台灣股市電子業為測試標的，發現此修正式濾嘴法則能夠有效的將平均年複合報酬率由13%提昇至23%。

The filter trading rule is a method for effectively trading stocks. The profitability of the filter rule has been extensively justified. Empirical results show that the transaction points suggested by the filter rule are not always profitable. This research aims to modify the filter trading rule to achieve higher profitability. The basic idea is to effectively locate a group of transaction points, which is with higher profitability and can be characterized by a few variables. The decision tree algorithm—a data mining method—is applied to solve the group-identification problem. The characteristics of the high profitable group are then used to develop an enhanced filter trading rule. The proposed trading rule is justified by using the stocks of the electronics industry in Taiwan as testing samples. Experiments show that the annual compound return rate would increase from 13% to 23% by such an enhancement.