應用基因演算法於動態投資組合保險中操作策略的最適化

Abstract

投資組合保險的觀念，在國外已風行多年，而今也正逐漸地在國內受到重視。其概念為將資金分配於不同之資產，並將此投資組合所面臨的損失風險控制於一定範圍之內，但卻又能參與市場上漲時的獲利。 市場上目前常用的動態投資組合保險之操作策略通常為CPPI與TIPP兩種，其簡單的參數設定與計算公式，使其在實務上易於使用。但其參數之設定卻常依賴於過去經驗值，並無一定之規則可循。本研究建立一動態投資組合保險之操作模型，以決定股票與債券兩者之間的資金權重比例與調整時機，在考慮實際交易成本之情況下，應用基因演算法於調整其槓桿乘數Multiplier以及最低交易調整門檻Tolerance，依投資者對風險的趨避程度予以最適化以求得單位風險報酬的最大化。 經由本研究的實證，可得以下之結論。以單位風險報酬來比較，無論在多頭或空頭走勢下，TIPP皆優於CPPI，而CPPI又優於大盤表現。而在不同的市場走勢與要保額度下，槓桿乘數以及最低交易調整門檻並無法得到一固定值，故本研究嘗試利用前半年市場資料所得出之最適值，實證於其連續後半年之市場中。結果在一連續的市場趨勢中，可得出較大盤為佳之表現；但在大盤向上反轉之階段，則劣於大盤。

The idea of portfolio insurance has been widely practiced abroad for years; while it is now gradually accepted over the inland. The critical point of the idea is to make capital allocated in various assets as for better risk management in additions to gain profit when economy is on bull market. The common strategies shown in dynamic portfolio insurance are "CPPI" and "TIPP", which practice is realized out of its simple calculation and parameters to set. On the contrary, for lacking of regulation to follow, the setting of parameter should attach to the figures of past experience. Besides, it makes use of GA to optimize the "Multiplier" and "Tolerance" considering of transaction cost in order to maximize Sharpe ratio according to investor's risk management plan. The following conclusions are summarized from the empirical result. Compared by Sharpe ratio, TIPP is superior to CPPI and CPPI outperform benchmark in either bull market or bear market. As the calculation of Multiplier and Tolerance fail to fix in various situation of market trend and desired floor, the experiment hereby tend to pick a best proper figures given by the previous half year of real market data, and testify the theory works in the later half year with real market practice. It shows that the model have better performance in continuous trend, but become worse than benchmark only when in the upside reversal point.