Investigating Agent Strategies within a Complex Adaptive System of Purchasing Agents for Estimating Attribute Relevance

Abstract

The authors constructed a complex adaptive system of purchasing agents, and have used this software to estimate the relevance of attributes in a data set. The simulation supports the creation of arbitrarily complex and diverse agent strategies. The effect of altering the proportion of two different agent strategies is investigated and reported, and contrasted with two non-CAS strategies based on information-gain variants