Intelligent Stock Trading Systems Using Fuzzy-Neural Networks and Evolutionary Programming Methods

摘要

The goal of this study was to analyze the possibilities of fuzzy neural networks and evolutionary programming methods for creating the human skill based stock trading systems. In stock exchange markets, the relationships between market variables are generally too complex to make rightful trading decisions and to earn stabile profits using classical system theory approach. On the other hand, there are a lot of trading experts-practicians that successfully trade stocks and achieve good results in the stock exchange markets. A useful technique for expert-knowledge extraction is the supervised learning methods, where human-experts actions are mapped using fuzzy-neural networks. In this paper we outline this procedure. Also we discuss the possibilities for improvement the proposed human skill based stock trading systems. An efficient biological system evolves slowly over the course of hundreds and thousands of generations of individuals. Later generations have more fit and are more capable than earlier ones. Similarly, we have used evolutionary techniques to "evolve" the fuzzy-neural network based stock trading system, which is capable to solve the stock trading task more efficiently. Proposed procedure was tested using virtual trading system that uses historical data from US stock markets. The first results confirmed the good opportunities of the proposed approach.