Summary
International Symposium on Nonlinear Theory and its Applications
2017
Session Number:A2L-D
Session:
Number:A2L-D-3
Predictability of Financial Market Indexes by Deep Neural Network
Tomoya Onizawa, Takehiro Suzuki, Tomoya Suzuki,
pp.166-169
Publication Date:2017/12/4
Online ISSN:2188-5079
DOI:10.34385/proc.29.A2L-D-3
PDF download (89.8KB)
Summary:
The present study investigates the predictability of financial markets such as TOPIX (Tokyo StockPrice Index) by using the stacked autoencoder for the dimensionality reduction of all the companies based on TOPIX and for the pre-training of deep neural network to learn complex movements of financial markets. Moreover, if neurons can be independent on a shallower layer, it would be better to use the naive Bayes classifier for the following layers. We perform some simulations with real stock data to investigate the above possibilities.