在投資市場上,投資者最常關注到挑選投資標的以及在市場風險下進行資產配置,因此本研究應用柔性運算的方法,發展兩階段的投資組合方法,第一階段透過資料包絡法(Data Envelopment Analysis)之CCR模式,發展出一套挑選基金模式;接著第二階段以提出兩個整合GA與PSO的演算法-GPSO1及GPSO2進行資產配置。 在實證研究中,本研究以台灣國內股票型基金為例。本研究主要提出以資料包絡法(Data Envelopment Analysis)來挑選基金作為本研究之投資組合標的;在資產配置上,以GPSO1、GPSO2、GA及PSO四種演算法建構投資組合並與大盤共五種投資組合分別比較Sharpe值。本研究之資料共計36個月,實證結果顯示所提之GPSO1與GPSO2投資組合之Sharpe值均優於GA、PSO及大盤分別所構成的投資組合,而其中又以GPSO1表現最佳。實證結果證明本研究提出的兩階段投資組合方法的確能幫助投資人穩健獲利。
In the investment market, investors concern greatly about investment target selection and asset allocation under the risk of investment market. Therefore, this research applies soft computing methods to develop a two-stage portfolio method. The first stage utilize a Data Envelopment Analysis (DEA) - CCR model to develop a model of funds selection, with the second stage integrates Genetic Algorithm (GA) and Particle Swarm Optimization Algorithm (PSO) to develop two method, GPSO1 and GPSO2, for assets allocation. The experiments were conducted by using the Taiwan domestic stock fund. In today, there are 36-month data. This study compares the proposed methods, GPSO1 and GPSO2, with GA, PSO, and market index based on Sharpe value. The evaluation results showed that GPSO-based methods outperform the other method for portfolio allocation. Especially GPSO1, it has the best performance. Thus, the proposed two-stage portfolio method really can help investors have stable profit.