Multi Agent System: three behavioural frameworks integrated into a GIS

In Multi Agent Systems applied to the social science simulations the construction of evolution rules applied to agents (individual agents and aggregated agents) follows three main approaches. A first approach is based on urban theories and rules derived from analytical models; a second one is based on “expert knowledge” and rules derive from the know-how of the modeler or from common sense; the last one is the more complex: rules derive directly from data. In this work we discuss adopted rules in some examples of Multi Agent Systems and describe three applications we developed. The first one uses the first mentioned approach, through a dynamic spatial interaction model integrated in a GIS software; the second one uses rules from “expert knowledge” and propose an integration between Geodatabase Data Model and MAS; the last one proposes a way to construct simple and readable rules starting from data and using Decision Tree Induction.