Etude du comportement humain grâce à la simulation multi-agents et aux méthodes de fouille de données temporelles

摘要

The simulation tools have already been tested for the study of many phenomena, but to simulate human behavior remains a challenge for both computer and human sciences. In this context, we have worked on the simulation of individual behavior and social interactions in groups of individuals to observe the emergence of social phenomena. To this end, we conduct an inter-disciplinary mix of artificial intelligence techniques, multi-agent systems, social studies and data mining spatio-temporal. This thesis presents a new model of emotional agent: the agent EFT (Emotion, Feeling, Temperament). Based on BDI architecture, our model integrates the OCC model emotional and behavioral model PerformanSe. We also present a concrete implementation of our model: the simulation of brain injured persons on a packaging line. We describe the modeling of the interactions of our system through AgentUML. Through methods of data mining spatio-temporal, we extract relevant information from data from the simulation.