Abstract

We present a comparative study of seven evolutionary algorithms (Generational Genetic, Elitist Genetic, Steady State Genetic, (μ/ρ, λ) Evolution Strategy, (μ/ρ + λ) Evolution Strategy, generational and elitist Covariance Matrix Adaptation) for automatic calibration of a constrained cellular automaton (CCA), whose performance are assessed in terms of two fitness metrics (based on Kappa statistics and Lee-Salee Index). Two variations of the CCA (one with 14 and one 27 parameters) were tested jointly with different number of time steps targeted by the calibration procedures. Besides offering some methodological suggestions for this kind of comparative analysis, the findings provide useful hints on the calibration algorithms to be expected to perform better in the application of cellular automata of sort for the simulation of land-use dynamics.

I documenti depositati in UnissResearch sono protetti dalle leggi che regolano il diritto d'autore