An intelligent fuzzy based system for market design agents

In this paper, we will present an intelligent fuzzy logic based system for market agents operating within the Trading Agent Competition's Market Design game. The objective of this competition is to design intelligent agents and systems that would be able to manage an electronic double auction market effectively in competition with other markets for market share and profit. Thus the market share and profit along with the market's transaction success rate constitute the agent's score in the game. Managing the market involves setting rules for accepting, matching and clearing offers as well as pricing transactions and imposing fees. In this paper, we will present our agent which is called MyFuzzy which employs a fuzzy logic based module to impose its fees in order to balance the profit and market share. The novelty of this approach lies in the fact that it makes minimal assumptions about the nature of the game environment by using fuzzy techniques to adapt the agent behaviour by recognizing the market context. This allowed our agent to achieve superior performance to the other existing techniques as it avoids factors that other agents overlook. Furthermore, we introduce novel modifications of existing pricing and accepting techniques to work in conjunction with our charging module. The proposed agent has been compared against the other agents (including the winners of the 2009 trading agents competition) where our agent has achieved superior performance. © 2010 IEEE.