Enhancing the Bees algorithm for global optimisation using search space manipulation

Binbakir, Turki (2021). Enhancing the Bees algorithm for global optimisation using search space manipulation. University of Birmingham. Ph.D.

Preview

Binbakir2021PhD.pdf
Text - Accepted Version
Available under License All rights reserved.
Download (3MB) | Preview

Abstract

The aim of this research is to improve the ability of the Bees Algorithm to tackle global optimisation problems. The Bees Algorithm was formulated and inspired by the foraging behaviour of honeybees. The proposed enhancements target the initialisation and global search stages of the algorithm. The reason for this is that the initialisation stage could save efforts by directing the search earlier towards the more promising areas of the search space, leading to a better optimised result. Targeting during the global search is due to the researcher’s belief that the neighbourhood search depends on it and any improvement will positively affect the neighbourhood search.
In this research, three enhancements were formulated based on the manipulation of the search space. The first enhancement (BAwSSR) involves continuous and gradual reduction of the search space with different scenarios that vary according to the starting point of reduction. The second enhancement (BADS) deals with the segmentation of search space into independent segments while using two sampling approaches to tackle a wide variety of problems. The third enhancement (BAOSS) also involves the segmentation of search space but divides it into independent segments to increase flexibility in handling a wider range of problems.
These proposed algorithms were tested on 24 benchmark functions with a broad range of characteristics. This test involves performance comparisons with the Quick Artificial Bee Colony (qABC) and the Standard Particle Swarm Optimisation 2011 (SPSO2011) algorithms. The obtained test data indicated noticeable improvements with an adequate level of stability over the original Bees Algorithm. The results were supported by the Mann–Whitney significance test, showing the improvements are statically significant for both accuracy and speed. Additionally, the proposed algorithms were tested on two engineering problems that included a comparison with a group of competitor algorithms. However, only the first proposed algorithm (BAwSSR) showed an obvious improvement. The other two algorithms (BADS) and (BAOSS) did not reveal any improvement.

Type of Work:

Thesis (Doctorates > Ph.D.)

Award Type:

Doctorates > Ph.D.

Supervisor(s):