On using ZCS in a Simulated Continuous Double-Auction Market

This paper presents results from ongoing investigations into the performance of the Michigan-style classifier system in a complex multi-agent environment. Using a simplified model of a continuous double-auction market place the use of ZCS as an adaptive economic trading agent is examined. It is shown that a number of small changes to the basic system greatly improves its performance, resulting in improvements in the overall efficiency of the market. It is also shown that the role of the rule-discovery component of the classifier system is particularly critical in such a closely-coupled multi-agent environment.