We propose modified hybridizations of genetic algorithms with some neighborhood search based meta-heuristics. In our hybrid algorithms, we consider gradually increasing probability for the application ofthe neighborhood search procedure on the best individuals as the number of iterations of the genetic algo-rithm increases. We implement the proposed hybrid algorithms and compare their performance withtwo other recently proposed hybrid algorithms which, in contrast, use the neighborhood search proce-dure on all the individuals of the population, two hybrid algorithms applying simulated annealing on thebest individual in the papulation in every iteration and three non-hybrid metaheuristic algorithms. Toinvestigate the effectiveness of the proposed algorithms, we apply the algorithms to our proposed fuzzybus terminal location problem models. The fuzzy model is considered to have fuzzy number of passen-gers corresponding to the nodes as well as fuzzy neighborhoods, together with preassigned lower andupper bounds for the number of required terminals. The algorithms are tested on a variety of randomlygenerated large scale fuzzy bus terminal location problems in with fuzzy cost coefficients. The fuzzyobjective is transformed into a crisp one by use of a ranking function. The computational experimentsdemonstrate the effectiveness of the proposed algorithms on large scale problems. Finally, to show theeffectiveness of our proposed hybridizations, we also make a comparative study of our algorithms oncrisp facility location and quadratic assignment test problems.

Keywords