Modified Spider Monkey Optimization based on Nelder–Mead method for global optimization

Abstract

This paper proposed a modified version of Spider Monkey Optimization (SMO) algorithm for solving global optimization problems. The traditional SMO consists of seven phases where each phase has its characteristics and tasks to be performed. However, the local leader phase (LLP), that is the second phase of the SMO has the most significant effect on the performance of the algorithm. In which if it does not has good exploration and exploitation capability then the SMO might stick at a local point. Therefore, the proposed modified version of SMO (that called SMONM) used the transformations of the Nelder–Mead (NM) method to improve the ability of LLP. The proposed SMONM algorithm contains the same number of phases of the traditional SMO except the LLP that has modified through using the reflection, expansion, and contraction transformations of the NM. These transformations of NM worked if there is no improvement in the fitness function value after the solution is updating using the original LLP. The performance of the proposed algorithm has compared with other four algorithms namely, original SMO, Artificial Bee Colony optimization, Biography Based Optimization and Particle Swarm Optimization. A set of experimental series is performed to evaluate the performance of the proposed algorithm using 23 standard benchmark functions, 15 composite functions, and three classical engineering problems. The preliminary results show that the modified version of SMO has excellent ability to avoid the limitations of the tradition SMO algorithm, as well as, it provides better results than the other comparative algorithms regarding performance measures.