A stock market decision support system with a hybrid evolutionary algorithm for many-core graphics processors

Abstract

This paper proposes a computational intelligence approach to stock market decision support systems based on a hybrid evolutionary algorithm with local search for many-core graphics processors. Trading decisions come from trading experts built on the basis of a set of specific trading rules analysing financial time series of recent stock price quotations. Constructing such trading experts is an optimization problem with a large and irregular search space that is solved by an evolutionary algorithm, based on Population-Based Incremental Learning, with additional local search. Using many-core graphics processors enables not only a reduction in the computing time, but also a combination of the optimization process with local search, which significantly improves solution qualities, without increasing the computing time. Experiments carried out on real data from the Paris Stock Exchange confirmed that the approach proposed outperforms the classic approach, in terms of the financial relevance of the investment strategies discovered as well as in terms of the computing time. © 2011 Springer-Verlag Berlin Heidelberg.