Multi-subroutines in Genetic Network Programming-Sarsa for trading rules on stock markets
Yang Yang, Shingo Mabu, Yunqing Gu, Kotaro Hirasawa · Society of Instrument and Control Engineers of Japan · 2011
This paper describes a decision-making model for creating trading rules on stock markets using a graph-based evolutionary algorithm named Genetic Network Programming-Sarsa (GNP-Sarsa) and multi-subroutines. The method is developed for discovering the repetitive subgraphs over the entire graph structure and modularizing them as subroutines, which results in substantially fastening the search by suppressing redundant search and results in reducing the overfitting leading to the improvement of the generalization capability. The following two are discussed: 1) varying the number of subroutine nodes in the main program and 2) varying the kind of subroutines to be generated. The experimental results on the stock markets show that the proposed method can generate more efficient and robust trading models and obtain much higher profits.