55 A Novel Particle Swarm Optimizer with Kriging Models
Download citation file:
In Particle Swarm Optimization (PSO), fitness functions play a very important role. However, if we only have some observation points and have no explicit fitness function of independent and dependent variables of these points, the applications of PSO will meet some obvious difficulties. In this paper, we used kriging models to address this problem. We used kriging models to construct fitness functions based on some known observation points and then according to the constructed fitness functions, PSO was adopted to perform the optimization operation. A novel algorithm combining PSO with kriging models was developed and three simulated cases were investigated. The results from these cases showed our algorithm successfully found the optimal points using only some known observation points.