Development of a fuzzy position controller for an autonomously trading agent

In this paper, we propose a fuzzy rule-based system that performs position control for futures trading. The fuzzy position controller enables a trading agent to balance in the amount of bought and sold futures. Each fuzzy if-then rule in the fuzzy rule-based system has weights. The weights are updated every settlement based on the result of the futures trade. That is, the fuzzy position controller performs an off-line learning. In our computer simulations, we evaluate the performance of a trading agent with the fuzzy position controller. We show through our computer simulations that the fuzzy rule-based system improves the trading performance in a virtual futures market.