Generating stock trading signals based on matching degree with extracted rules by genetic network programming

When action rules of agents are created by evolutionary computation, it generally aims to create optimal individual which represents optimal rules. On the other hand, genetic network programming (GNP) with rule accumulation extracts a large number of rules throughout the generations and store them in the rule pools. In other words, the individuals of GNP with rule accumulation are regarded as rule generators which are evaluated by fitness function every generation. In this paper, GNP with rule accumulation is applied to generating buying and selling rules in a stock market, and a large number of rules are extracted by the individuals which contribute to the fitness in the training period. Then, the trading in the testing period is carried out using the extracted rules and the profits of the testing results are evaluated.