Parameterization of a metapopulation model: an empirical comparison of several different genetic algorithms, simulated annealing and tabu search

Abstract

Analysis of metapopulation dynamics is currently of great interest in population biology and in conservation biology. In this study a metapopulation model is augmented with external environmental factors, which are modelled by a group of polynomials. The parameter estimation of the extended model is attempted with three methods of global optimization, simulated annealing (SA), tabu search (TS) and genetic algorithms (GA). GA variants tested include a binary coded implementation, several floating point implementations such as the breeder GA (BGA), and a simulated diffusion model parallel implementation. The BGA produced the most consistent convergence whereas SA eventually produced the lowest value of the objective function.