A Genetic Algorithm Approach to Supporting a Three-Staged Portfolio Construction Process

• Main Authors: Min-Jiun Lin, 林敏鈞
• Other Authors: Shin-Yuan Hung
• Format: Others
• Language: zh-TW
• Published: 2000
• Online Access: http://ndltd.ncl.edu.tw/handle/95360030572345179097

碩士 === 國立中正大學 === 資訊管理學系 === 88 === The whole process of portfolio construction can be separated into three steps, including choosing the right assets, predicting the future prices of the assets, and constructing the portfolio. The research pursues an improvement in the last two steps. We modified the approach developed by Hung et al. (1996) (ANN-Case Model) and tried to find out the globally optimal portfolio by integrating the genetic algorithm. The new approach (ANN-GA Model) firstly uses the arbitrage pricing theory to price risk assets. This is followed by using the artificial neural network to predict the future trend of each risk factor. Finally, the genetic algorithm is applied to optimize the portfolio. In order to verify the effectiveness of the new approach, we adopted two time periods for evaluation. The first time period is the same as prior research for comparison and the other time period is recent data to reflect the current trend of the Taiwan stock market. The empirical findings show that the ANN-GA Model beats both benchmarks (i.e. TSEWPI and MSCI Taiwan stock index) and outperforms the ANN-Case Model. Further discussions are also provided as well.