Generating Stock Trading Rules Using Genetic Network Programming with Flag Nodes and Adjustment of Importance Indexes

摘要

Genetic network programming (GNP) is an evolutionary algorithm which represents its solutions by means of graph structures. Since GNP can create quite compact programs and has an implicit memory function, it works especially well in dynamic environments. In addition, a study of the creation of trading rules on stock markets using GNP with Importance Index (GNP-IMX) has been made. The IMX is one of the criteria for decision making. However, the values of the IMXs must be determined by experience and knowledge. Therefore, in this paper, IMXs are adjusted appropriately during stock trading in order to predict the rise and fall of the stocks. In addition, newly defined flag nodes are introduced into GNP, which makes it possible to appropriately judge the current situation of the stock prices, and also contributes to the use of many kinds of nodes in the GNP program. In a simulation, programs were evolved using the stock prices of 20 companies. Then the generalization ability of the method was tested and compared with GNP without flag nodes, GNP without IMX adjustment, and Buy & Hold.