應用類神經網路建立股票型基金市場績效預測模型-以整合技術指標與門檻值設定之分析為例

Abstract

投資理財是門不容小覷的學問，投資者須解讀眾多訊息，才可趨利避險。共同基金是一種「共聚資金、共擔風險、共同分享投資利潤」的投資方法，其倚靠著風險小、有專業人士操盤管理、流通性大等好處，逐漸成為一般投資人喜好的投資管道。但自從2008年發生全球金融海嘯後，全球經濟情勢變得詭譎多變，不僅股市的表現更加難以掌握，連帶著基金市場也不再是長期持有就可以穩定獲利，基金投資也變成須隨時注意市場的變化。 本研究挑選國內基金市場中155檔股票型基金，將基金與股市的報酬率關係，搭配長短期的技術指標，透過人工智慧中的倒傳遞類神經網路，進行趨勢行為學習。經由類神經網路強大的非線性搜尋能力，嘗試從中尋找出隱含的行為，進而建立一個股票型基金市場投資報酬績效預測模型，並依此模型進行投資交易模擬。 由實證結果顯示，採用基金與股市的報酬率關係所建立的模型，在預測及獲利能力上，皆優於使用基金報酬率所建立的模型。證實運用本研究模型進行基金投資將可為投資人帶來平穩的獲利；實驗過程中亦發現依據模型的門檻值設定所得之買入時機點進行投資，其投資報酬相對較佳。

Finance and investment is the knowledge which should not be underestimated. It requires investors to interpret a plenty of information to make it profitable. Mutual fund is an approach of gathering fund with shared risk and return. It becomes a preferred investment pipeline because of its lower risk, professional manipulation and better capital liquidity. But, since the global financial crisis in 2008, the global economic situation has become so volatile. Not only the stock market is more difficult to handled, but the fund market is also no longer profitable for sure. It requires attention to the fund market change anytime. The correlation between fund and stock market incorporated with long term and short term indexes for the selected 155 domestic stock funds was analyzed through artificial intelligence in the neural network. With the powerful search capability of non-linear neural network, the implicated behaviors were revealed and a model of stock funds performance prediction was hereby established. Simulative transactions were conducted based on this model. Better forecast accuracy and profitability were demonstrated with the model created by the correlation of fund and stock market compared with the one built by the fund profitability according to this study. It was proven by this study that the model could lead to steady and sustainable profit. Furthermore, it was also found that the return was relatively better if funds were purchased at the timing of model suggested.