應用動量權重及灰預測傅立葉殘差修正模型於台灣股票市場電子股之投資組合策略

自Markowitz於1952年提出平均數－變異數投資組合模型架構以來，如何在MV模型之效率前緣下、建立最佳投資組合，一直是各界研究及探討的議題。效率前緣之投資組合通常都具有高報酬伴隨高風險以及低報酬伴隨低風險的特性，使得投資者必須在期望報酬率與風險之間取捨而造成投資決策進退兩難，其計算過程複雜，當標的物數量很多的時候，亦可能造成無法即時求出最佳解的情形。因此，本研究以人工智慧方法提出新模型，以合理及簡便的方法企圖解決問題。除此之外，有鑒於模型特性以及投入資金水準不同，本研究並與過去提出之模型做比較，分析模型的過度反應及不同資金水準下的表現，以便在趨近現實之情形下，反應出模型特性與優劣，做為投資者建立投資組合之參考。實驗結果顯示，多數投資組合模型存有過度反應的現象，且各種不同模型於不同資金水準下，表現並不相同，故模型之建立需考慮過度反應及資金水準之因素，故本研究提出動態投資線，在不同資金水準下，投資人可選擇最佳投資組合模型，以解決此問題。

關鍵字

灰色理論；傳立葉殘差修正；動量生命週期假說；投資組合；過度反應

並列摘要

Since Markowitz proposed mean-variance portfolio model in 1952, portfolio optimization, based on an efficient frontier, has continuously been discussed. Normally, the portfolios on the efficient frontier show a positive correlation between risk and return. Higher returns are always companied with higher risks, and vise versa. It leads to a dilemma for investors to make their decisions. And moreover, when there are too many securities or asset classes in the pool, it's difficult to obtain a precise solution under some limit of time. This research applies artificial intelligence to offer a rational and convenient solution on portfolio optimization. Besides, this research compares a few models by analyzing the performances under different level of over-reaction and size of investing capitals. The results reflect the characteristics and the pros and cons of the models in more realized circumstances. It provides a good reference for portfolio building. The result of this research shows that over-reaction appears in most models. And the performances are quite different according to different capital levels. So we provide a dynamic invest line to express the optimization-portfolio model in the different capital levels.