Akademik Veri Yönetim Sistemi
Tezin Türü: Doktora
Tezin Yürütüldüğü Kurum: Marmara Üniversitesi, Fen Bilimleri Enstitüsü, Türkiye
Tezin Onay Tarihi: 2024
Tezin Dili: İngilizce
Öğrenci: RAMAZAN ALGIN
Danışman: Ali Fuat Alkaya
Açık Arşiv Koleksiyonu: AVESİS Açık Erişim Koleksiyonu
Özet:
| Meta-heuristics are commonly used solution techniques for combinatorial optimization problems. They are preferred to exact algorithms when optimum solutions are sought on large problem instances. However, for instances of high complexity or large-scale problems, heuristics or meta-heuristics may not be sufficient to achieve satisfactory results. For this reason, especially during the last three decades, researchers have been trying to find new techniques that provide better performance. One of the fields that researchers are focusing on is hybridizing meta-heuristics. Hybrid meta-heuristics are generally obtained by combining the power of two or more meta-heuristics or by placing a local search heuristic within a meta-heuristic. Developing well-performing methods for problems described in this thesis is still an open issue and hybridization of existing methods to obtain better results is a widely applied method. Therefore, in this thesis, developing hybrid techniques for problems in different domains is studied. This thesis focuses on four problems from both discrete and continuous domains; feature selection problem, obstacle neutralization problem, quadratic assignment problem and continuous functions. We proposed three new hybrid techniques by combining migrating birds optimization with simulated annealing, simultaneous perturbation stochastic approximation algorithms and reinforcement learning technique. These three hybrid techniques and several state-of-the-art meta-heuristics are successfully applied to the four problems and their performances are compared on the problem instances. Computational experiment results are presented to show effectiveness of the developed hybrid techniques for each problem. |