An evolutionary multi-agent system for simulating e-commerce markets

This paper describes a multi-agent system used for simulating a generic e-trading market. The simulated market is imagined to be on the Internet. Four different types of agents are used in our system: clients, producers, distributors and mediators - for clients and producers. The simulation starts with an initially given market which evolves under various assumptions about mediators. Mediation services support the following aspects: brokerage, billing, delivery and connectivity. Each mediator has a profile based on its degree of greediness, effort and speed. The evolution of the system is controlled by the individual fitness of the mediator agents. Those that profit from their negotiations stay alive, while others die. The most successful mediators are cloned and populate the system. Thus, the system evolution helps with identifying the appropriate mediator profiles. Internal implementation choices are discussed and analyzed. Results from an experiment using a typical market scenario are presented to illustrate the system in use.