應用基因演算法及自組織映射圖神經網路對外資在台股指數期貨持有成本之分析與大盤走勢行為知識發現

Abstract

自股票市場開放外資投資以來，大戶坑殺散戶之情形時有所聞；三大法人（包含外資）挾以資訊優勢，擁有良好的擇時能力，而資訊不對稱下，散戶總是追漲殺跌的最大受害者。過去文獻多探討三大法人在現貨市場之行為，較少探討其在期貨市場之行為，而在針對期貨未平倉量之研究，均無針對外資於期貨的佈局及持有成本進行分析。因此本研究將利用台灣期交易所於2008年4月7日才公開的三大法人每日未平倉部位資料，進行外資於台指期貨市場交易行為之研究。 近年來，人工智慧方法學在財務金融領域上之應用蓬勃發展，其中基因演算法被公認為穩健有效率的最佳化方法。因此本研究使用基因演算法架構，期能求出外資於台指期貨的佈局與持有成本，並將求得之外資損益連同籌碼面指標輸入自組織映射圖神經網路，進行分群，提出一個具有趨勢預測能力之人工智慧模式。 本研究取樣資料來源為台灣期交所日內交易及證交所大盤日交易資料，從2007年07月02日至2008年12月31日總共377個交易日資料。實證結果發現外資在期貨市場的未平倉部位變化與其持有成本的確能提供有用的訊息協助判斷大盤的趨勢。

Since the stock market investment is available for foreign investors, it occurs that the rich family of the stock takes advantage of the private investor frequently. The three major institutional investors (including foreign investor) take advantage of the information and hold a good ability of timing, but under the information asymmetric, the private investor is always the victim. The past reference usually talked about the behavior of the three major institutional investors in the spot market, but seldom discussed the behavior in the futures. In the open interest, there is no analysis that the market strategy and the cost of carry in futures of foreign investor. As a result, in this research, we use the data of open interest of the three major institutional investors in Taiwan Future Exchange reported on April 7th, 2008 to do the research on the trade by the foreign investor in the futures. Recently, the application of the artificial intelligence approaches in the field of finance is developed enormously, the genetic algorithm especially. Therefore, in this research, we use the genetic algorithm to get the market strategy and the cost of carry of the foreign investors in Taiwan Futures. And then importing the profit and loss of the foreign investors with institutional analysis to the self organizing map neural network, bring up an artificial intelligent module of predictive ability. In this research, the data are from daily market report in Taiwan Future Exchange and Taiwan Stock Exchange including 377 trading dates between July 2nd 2007 and December 31st 2008. In the experiment, we find that the change of the open interest and cost of carry in futures of the foreign investors can give the helpful information to decide the trend of market index.