Record Details

Using improved BPN/Cauchy machine and genetic algorithms to build an efficient neural network and to forecast Taiwanese electronic stock indexes

Journal of Financial Studies

View Archive Info
 
 
Field Value
 
Title Using improved BPN/Cauchy machine and genetic
algorithms to build an efficient neural network and
to forecast Taiwanese electronic stock indexes
 
Creator Deng-Yiv Chiu
Tien-Shang Chen
Ya-Chen Pan
 
Description In recent years, artificial neural network (ANN) was used to forecast stock prices in many research efforts.
Some of those efforts lacked a criterion rule to choose input variables and to achieve a significant network
architecture. Some of them used trial-and-error to find a better network architecture in some range.
Therefore, in this research, we use genetic algorithms (GA), a neural network and an improved
Back-Propagation Network/Cauchy (BPN/Cauchy) machine as the learning algorithms to train the network.
The goal is to prevent the network from local optimum and achieve a satisfactory result efficiently. For this
kind of network, the training period is long when it is applied to a complex model. We modify the method to
train neural network to obtain balance between time consumption and performance. We expect to improve the
ability of the network prediction model through our proposed methods. Finally, our proposed method is applied
to forecast Taiwanese electronic stock indexes. The results of the experiment reflect that our proposed method
is feasible for building an efficient neural network.

Key words: Artificial Neural Network, Genetic Algorithm, Electronic Stock Indexes, BPN/Cauchy Machine
 
Publisher Journal of Financial Studies
財務金èžå­¸åˆŠ
 
Date 2011-05-27
 
Type —
 
Format application/pdf
 
Identifier http://www.jfs.org.tw/index.php/jfs/article/view/2011126
 
Source Journal of Financial Studies; Vol 16, No 4 (2008); 213
財務金èžå­¸åˆŠ; Vol 16, No 4 (2008); 213
 
Language —