Improved Dragonfly Algorithm with Neighbourhood Structures

Abstract

Dragonfly algorithm is a recently proposed optimization algorithm inspired on the static and dynamic swarming behaviour of dragonflies. Because of its simplicity and effectiveness, DA has received interest of specialists from various fields. Premature convergence and local optima is an issue in Dragonfly Algorithm. Improved Dragonfly Algorithm with Neighbourhood Structures (IDANS) is proposed to overcome this drawback. Dragonfly Algorithm with Neighborhood structures utilizes candidate solutions in an iterative and intuitive process to discover promising areas in a search space. IDANS is then initialized with best value of dragonfly algorithm to further explore the search space. In order to improve the efficiency of IDANS, Neighbourhood structures such as Euclidean, Manhattan and Chebyshev are chosen to implement these structures on IDANS to obtain best results. The proposed method avoids local optima to achieve global optimal solutions. The Efficiency of the IDANS is validated by testing on benchmark functions and classical engineering problem called Gear train design problem. A comparative performance analysis between IDANS and other powerful optimization algorithms have been carried out and the results shows that IDANS gives better performance than Dragonfly algorithm. Moreover it gives competitive results in terms of convergence and accuracy when compared with other algorithms in the literature.