Knowledge-based decision support systems for dealing nikkel-225 by soft computing techniques

摘要

This article introduces a novel way in dealing Nikkei-225 (a stock price index in Japanese stock market based on the Dow formula). Due to the uncertain nature of the stock market, the traditional approaches such as regression analysis seem to be not so effective any more, so we prefer to use the soft computing techniques including the combination of neural networks (NNs) and genetic algorithms (GAs). After considering several effective impacts on the changeful stock market, we select some as the inputs of NN. And then we tried both the fixed and the shift methods to train the NN in order to learn the patterns of historical data. Simulation shows that: Shift method is more adaptable, whose prediction of Nikkei-225 four weeks in the future presents a high coincidence ratio with the real Nikkei-225 four weeks in the future, and these two values are highly correlated. We construct a Decision Support System (DSS) based on this prediction and further use GAs to optimize the parameters and the rules of dealing strategy. The final DSS can get a considerably big return under the practical condition that the tax and processing fee has been subtracted every time in the dealing.