Evolutionary models for agent-based complex behavior modeling

Abstract

In this chapter, the essentials of genetic algorithm (GA) following the footsteps of Turing are introduced. We introduce the connection between Turing's early ideas of organized machines and modern evolutionary computation. We mainly discuss the GA applications to adaptive complex system modeling. We study the agent-based market where collective behaviors are regarded as aggregations of individual behaviors. A complex collective behavior can be decomposed into aggregations of several groups agents following different game theoretic strategies. Complexity emerges from the collaboration and competition of these agents. The parameters governing agent behaviors can be optimized by GA to predict future collective behaviors based on history data. GA can also be used in designing market mechanisms by optimizing agent behavior parameters to obtain the most efficient market. Experimental results show the effectiveness of both models. Using evolutionary models may help us to gain some more insights in understanding the complex adaptive systems.