Experiments in Predicting Financial Time Series using Modular Neural Network Architectures
Abstract
Forecasting of stock returns and stock market behavior has received special attention as application of smart engineering systems. In previous experiments, this problem lias been most commonly attacked using multi-layer perceptrons or Probabilistic Neural Networks. An adaptive modular network, as developed in this paper, weighs the decisions of several architectures for a final output depending on the input pattern and can thus combine the advantages of both architectures. Test runs show that the proposed architecture is clearly able to outperform the index, although results are not as good as in previous experiments. The network achieves an average final portfolio value about 2% higher than the index.
Recommended Citation
K. Schierholt and C. H. Dagli, "Experiments in Predicting Financial Time Series using Modular Neural Network Architectures," Intelligent Engineering Systems Through Artificial Neural Networks, vol. 6, pp. 769 - 774, Dec 1996.
Department(s)
Engineering Management and Systems Engineering
Document Type
Article - Journal
Document Version
Citation
File Type
text
Language(s)
English
Rights
© 2024 The Authors, All rights reserved.
Publication Date
01 Dec 1996
