資訊管理學報
陳美支;黃銘嘉;陳安斌;
頁: 67-89
日期: 2006/10
摘要: 近年乘國人投資理財的風氣漸盛，金融商品日益多元化，面對台灣淺碟式經濟股市經常大起大落，如何降低投資風險，避免鉅額損失，是投資者尋求的投資之道。投資組合保險(Portfolio Insurance)的概念是籍由付出少許的保險費用，以鎖定整個投資組合所面臨價格下跌時之風險，將損失控制於一定範圍之內，市場上漲時卻又不失參與獲利機會。時间不變性投資組合保險策略（Time Invariant Portfolio Protection，簡稱TIPP）是投資組合保險常用的策略之一，其中槓桿乘數(Multiple)及最低交易調整門檻值(Tolerance)等參數，全憑投資者經驗而設定，然而股市的變化莫測，一般投資人極難掌握其變化趨勢。本研究嘗試以時间不變性投資組合保險策略為基礎，利用多重分類元系統(Multi-Agent Extended Classifier System)，動態調整槓桿乘數、最低交易調整門檻值。輸入因子採用移動平均線(MA)、隨機指標(KD)、平滑異同移動平均線(MACD)、相對強弱指標(RSI)等技術指標，輸出因子則為隔日趨勢判斷、槓桿乘數及最低交易調整門檻值。本研究並以台灣股價加權指數(TAIEX)為研究標的，實證結果並與標準的TIPP策略比較，多重分類元系統模型確實有較佳的獲利表現。
關鍵字: 多重分類元系統;時间不變性投資組合保險;動態投資組合保險;
Abstract: This study attempts to investigate a novel hybrid scheme of Genetic Algorithm(GA) and Extended Classifier System(XCS)to explore Time Invariant Portfolio Protection (TIPP)policy. The multi-agent XCS-based approach dynamically optimizes Multiple and Tolerance variables concerned as the important parameters of TIPP. The proposed multi-agent classifier system analyzes the optimized parameters and stock technical indicators such as Moving Average(MA), Moving Average Convergence and Divergence (MACD), Stochastic Line(KD), Relation Strength Index(RSI), and volume to predict TAIEX trend and recommend Multiple and Tolerance. An evaluation of the proposed scheme is conducted based on 80% insurance ratios and periods of TAIEX from 1991 to 2004. The evaluation revealed that the multi-agent XCS-based approach dynamically identified adequate Multiple and Tolerance parameters and predicted market trend effectively. The evaluation results also suggested the proposed model outperformed traditional TIPP policy.
Keywords: Multi-Agent Classifier System;Time Invariant Portfolio Protection;Dynamic Portfolio Insurance;
瀏覽次數: 9288 下載次數: 119
引用 導入Endnote
頁: 67-89
日期: 2006/10
摘要: 近年乘國人投資理財的風氣漸盛，金融商品日益多元化，面對台灣淺碟式經濟股市經常大起大落，如何降低投資風險，避免鉅額損失，是投資者尋求的投資之道。投資組合保險(Portfolio Insurance)的概念是籍由付出少許的保險費用，以鎖定整個投資組合所面臨價格下跌時之風險，將損失控制於一定範圍之內，市場上漲時卻又不失參與獲利機會。時间不變性投資組合保險策略（Time Invariant Portfolio Protection，簡稱TIPP）是投資組合保險常用的策略之一，其中槓桿乘數(Multiple)及最低交易調整門檻值(Tolerance)等參數，全憑投資者經驗而設定，然而股市的變化莫測，一般投資人極難掌握其變化趨勢。本研究嘗試以時间不變性投資組合保險策略為基礎，利用多重分類元系統(Multi-Agent Extended Classifier System)，動態調整槓桿乘數、最低交易調整門檻值。輸入因子採用移動平均線(MA)、隨機指標(KD)、平滑異同移動平均線(MACD)、相對強弱指標(RSI)等技術指標，輸出因子則為隔日趨勢判斷、槓桿乘數及最低交易調整門檻值。本研究並以台灣股價加權指數(TAIEX)為研究標的，實證結果並與標準的TIPP策略比較，多重分類元系統模型確實有較佳的獲利表現。
關鍵字: 多重分類元系統;時间不變性投資組合保險;動態投資組合保險;
A Dynamic Portfolio Insurance Strategy Using Multi-Agent Extended Classifier System
Abstract: This study attempts to investigate a novel hybrid scheme of Genetic Algorithm(GA) and Extended Classifier System(XCS)to explore Time Invariant Portfolio Protection (TIPP)policy. The multi-agent XCS-based approach dynamically optimizes Multiple and Tolerance variables concerned as the important parameters of TIPP. The proposed multi-agent classifier system analyzes the optimized parameters and stock technical indicators such as Moving Average(MA), Moving Average Convergence and Divergence (MACD), Stochastic Line(KD), Relation Strength Index(RSI), and volume to predict TAIEX trend and recommend Multiple and Tolerance. An evaluation of the proposed scheme is conducted based on 80% insurance ratios and periods of TAIEX from 1991 to 2004. The evaluation revealed that the multi-agent XCS-based approach dynamically identified adequate Multiple and Tolerance parameters and predicted market trend effectively. The evaluation results also suggested the proposed model outperformed traditional TIPP policy.
Keywords: Multi-Agent Classifier System;Time Invariant Portfolio Protection;Dynamic Portfolio Insurance;
瀏覽次數: 9288 下載次數: 119
引用 導入Endnote