Parameter control for hybrid metaheuristic algorithms

A hybrid metaheuristic is defined as combining a metaheuristic with at least one algorithm. If parameter values are determined at run time of the algorithm, it is called parameter control. The aim of this thesis is to develop novel optimization methods by using hybrid metaheuristics and parameter control methods together. In this context, methods have been developed in two main optimization domains, namely continuous optimization and combinatorial optimization. For continuous optimization, bound constrained problems have been studied, and ABC, DE, and PSO metaheuristics, which are widely used in this area, were hybridized with high-level and relay principle of operation. As a parameter, the running order of these metaheuristics in this developed hybrid algorithm were determined by the adaptive parameter control based on the UCB selection algorithm. The experimental study shows the feasibility of metaheuristic selection with parameter control at algorithm level. For combinatiorial optimization domain, an algorithm that hybridizes ILS and VND metaheuristics with low-level and teamwork principle of operation was developed for Capacitated Vehicle Routing Problem. In this algorithm, an adaptive parameter control has been applied for acceptance function. The experimental study shows that the usage of parameter control in the proposed method is statistically significant. In addition, measured run-time and mean error rates are comparable to state-of-the-art algorithms in the literature.