Genetic network programming with credit
Wei Xu, Lutao Wang, Shingo Mabu, Kotaro Hirasawa · Society of Instrument and Control Engineers of Japan · 2012
The purpose of this paper is to study how to improve the performance of Genetic Network Programming(GNP) by adding a credit mechanism and its application to trades on stock markets. GNP is one of the evolutionary computations having directed graph structures. Because of this unique structures, GNP could be more effective for gaining profits in stock markets than the other methods. However, the structure of GNP has a high possibility to contain redundant nodes. These nodes will make negative effects to the whole structure. The proposed method will use a new branch named credit branch for making those redundant nodes pruned. Pruning is determined by both evolution and Sarsa-Learning of GNP. This algorithm can efficiently improve the performance of GNP. When the credit branch is chosen, the node will be skipped and the content will also be neglected. In the simulations, the stock prices of different brands from 2001 to 2004 are used to test the effectiveness. The results show that the proposed approach can provide better result than conventional GNP without credit branch.