Dissociation Patterns in Evolving Populations

Abstract

The recent explosion and availability of mobility based technologies such as geographic information systems, cell phones equipped with built‐in GPS, among others, are a valuable source of spatio‐temporal data. However, only recently there have been works focused on identifying movement patterns in groups of moving entities. We focus on a particular movement pattern: dissociation. A dissociation pattern occurs when an entity that was once associated to a population, eventually separated from it and subsequently reintegrated it again. The backwarding and forwarding patterns are a type of dissociation where an entity stays behind or ahead of another entity, respectively. Dissociation really is a diversity generator, so instead avoiding it, taking advantage could be better to prevent premature convergence in evolutionary algorithms. In this work, we present formal mathematical definitions for these patterns. A discussion of how to use dissociation patterns as a mean to preserve diversity in evolutionary algorithms is also shown.