Modified Cuckoo Search Algorithm Using Sigmoid Decreasing Inertia Weight for Global Optimization

Cuckoo Search (CS) is an evolutionary computational (EC) algorithm inspired by the behavior of a cuckoo bird, introduced by Yang and Deb in 2009 to solve various engineering-intensive optimization problems. However, this metaheuristic algorithm, CS, still suffers from premature convergence, mainly due to multimodal problems leading to local trap problems. This research introduces an adaptive swarm-based optimization approach to the CS algorithm, using the sigmoid decreasing inertia weight (DIW), which produces the modified Cuckoo Search using decreasing inertia weight (MCS-DIW) algorithm to tackle local trap problems. The paper shows that the proposed MCS-DIW depicts a better-controlled mechanism by adding the DIW with Lévy flight, for balanced exploration and exploitation in the global search domain. Moreover, this study presents an inclusive, experimental analysis of the widely used set of standardized benchmark test problems released by the Institute of Electrical and Electronics Engineers (IEEE) Congress on Evolutionary Computation (CEC) benchmark along with selected mathematical test functions to assess the performance of the MCS algorithm. The MCS-DIW algorithm is compared with other swarm intelligence (SI) algorithms to validate, including the original CS algorithm, Whale Optimization Algorithm (WOA), Sine Cosine Algorithm (SCA), and Search Sparrow Algorithm (SSA). The compiled simulation findings showed that the modified proposed CS algorithm, in most cases, performed better in attaining a low mean global minimum value, high convergence rate, and low central processing unit (CPU) processing time compared to other counterparts. The dynamic adjustment of inertia weight enhances optimization performance with an initial high inertia weight (e.g., 0.9) and promotes exploration, gradually decreasing to 0.2 for better exploitation. This proposed MCS-DIW approach provides faster convergence and has been proven to mitigate premature convergence. It reduces the number of iterations by 30-40% and achieves lower fitness values (e.g., 10-2) than static inertia weight, which often stabilizes at higher values (e.g., 10-1). In sum, the proposed MCS-DIW algorithm is proven to mitigate the local trap problems via an improved capability in searching for the global optimum.