A Fuzzy Rule-Based Trading Agent: Analysis and Knowledge Extraction

Abstract

In this paper, we show how a fuzzy rule-based system is developed for trading in a futures market. By our fuzzy rule-based system, an agent determines whether it should buy a futures spot or not based on the time series of both the spot and the futures prices. The fuzzy rule-based system is fine-tuned so that the amount of profit is maximized. Since a fuzzy system is used as a decision making tool, the decision making process by the learning agent can be linguistically interpreted. The performance of the fuzzy rule-based system is evaluated in a virtual stock market. We also try to extract a knowledge base from the fuzzy rule-based system after it is fully trained. Statistical test shows the effectiveness of the extracted knowledge as a human decision support.