The experimental study of population-based parameter optimization algorithms on rule-based ecological modelling

Abstract

This study investigates six population-based algorithms for the parameter optimization (PO) within the hybrid methodology developed for modelling algal abundance by rule-based models. These PO algorithms include: (1) Hill Climbing (2) Simulated Annealing (3) Genetic Algorithm (4) Differential Evolution (5) Covariance Matrix Adaptation Evolution Strategy and (6) Estimation of Distribution Algorithm. The effectiveness of algorithms is tested on the Cylindrospermopsis abundance data from Wivenhoe Reservoir in Queensland (Australia). We provide a systematic analysis and comparison of different parameter optimization algorithms as well as the resulting predictive rule models.