Zeki Sistemler Teori ve Uygulamaları Dergisi

Yazarlar: Rafet DURGUT, İlim Betül YAVUZ, Mehmet AYDİN

Konular:Bilgisayar Bilimleri, Yapay Zeka, Endüstri Mühendisliği

DOI:10.38016/jista.854584

Anahtar Kelimeler:Artificial Bee Colony,Adaptive Operator Selection,Set Union Knapsack Problem

Özet: Metaheuristic and swarm intelligence algorithms have been utilised to solve optimization problems with NP-Hard nature providing approximate solutions for a long time. Especially in the case of combinatorial and binary problems, operator functions embedded in the algorithms to generate neighboring solutions play a crucial role in the success of the algorithms while each operator imposes limitations upon the diversity of the search. In order to escape of such limitations and improve the diversity, multiple operators are preferred to use through a selection scheme instead of a single operator. However, the nature of selection scheme whereby operators are opted out also matters for higher efficiency, where a number of operator selection schemes have been used to solve various combinatorial problems with different swarm intelligence and metaheuristic algorithms before. In this paper, set union knapsack problems are, first time, solved with a binary artificial bee colony algorithm embedded with multiple operators selected through alternative operator selection schemes. Different credit assignment approaches, different sliding window lengths and parameter configurations are tested for the proposed method. The characteristics of the selection schemes are studied and the best performing one is suggested using a comprehensive experimentation over 30 benchmark problems. The study concludes a particular variant of binary artificial bee colony algorithm with a successful selection scheme.

---