Using Deep Neural Networks for Stock Market Data Forecasting: An Effectiveness Comparative Study

Abstract:

Stock market value forecasting has been a challenge, because data are massive, complex, non-linear and noised. Nevertheless, some deep learning promising techniques can be reviewed in technical literature. Using S&P500 historical data as a case study, this work proposes the following approach: (i) NARX and Back Propagation Neural Networks are selected and trained for representing Index data; (ii) A sliding window technique for Index value forecasting is defined and tested; and, (iii) An effectiveness comparison is performed. The results suggest the best model for representing and forecasting S&P500 Index data. Thus, the academics can revise a new experience in data analysis; and practitioners will have an approach concerning the forecasting calculation in the stock market.

Año de publicación:

2020

Keywords:

• deep learning
• S&P500 index
• stock market
• forecasting
• Neural networks

Fuente:

scopus