Applying Neural Networks to Discover Homogeneous Stock Groups

碩士 === 國立中正大學 === 資訊管理所 === 94 === This study applies neural networks to discover homogeneous stock groups. Self-Organizing Map (SOM), an unsupervised learning neural network model, is chosen to group homogeneous stocks. The grouping results were empirically tested and compared with those of the clu...

Full description

Bibliographic Details
Main Authors: Yuen-Han Chao, 趙元瀚
Other Authors: Shin-Yuan Hung
Format: Others
Language:zh-TW
Published: 2006
Online Access:http://ndltd.ncl.edu.tw/handle/07644008389266794251
Description
Summary:碩士 === 國立中正大學 === 資訊管理所 === 94 === This study applies neural networks to discover homogeneous stock groups. Self-Organizing Map (SOM), an unsupervised learning neural network model, is chosen to group homogeneous stocks. The grouping results were empirically tested and compared with those of the cluster analysis (CA) grouping method. Data in the Taiwan stock market from January of 1997 to December of 2005 was used to empirically test the effectiveness of those two grouping methods. The empirical results indicate that: (1) the SOM grouping method outperforms the CA grouping method in terms of portfolio returns; (2) the portfolio risks of the SOM grouping method are lower than those of the CA grouping method. The results demonstrate the effectiveness of the SOM grouping method.