<p>Many existing community detection algorithms for complex networks are based on modularity optimization but are limited in their ability to accurately assign community membership to individual nodes. Thus, this paper proposes the I-GWObCD algorithm for community detection in complex networks. It introduces a novel boundary local search strategy that incorporates a discrete optimizer, specifically the discrete improved grey wolf optimizer, which is among the most effective optimizers. The proposed algorithm effectively redefines the solution vector representation, movement strategy, and distances between wolves to solve the discrete community detection problem without prior knowledge. The proposed boundary local search modifies nodes on community boundaries to improve the extracted communities. A distance matrix was also used to simulate the construction of the neighborhood of a solution vector, as in I-GWO. The fitness function of the proposed I-GWObCD was also defined based on modularity, and the algorithm’s performance was evaluated on artificial and real-world network datasets in terms of modularity and NMI. The results were then analyzed using eight recent and advanced community detection algorithms. Moreover, the Friedman test was employed to analyze the results statistically. The experimental findings demonstrate that the proposed I-GWObCD effectively identifies higher-quality communities compared to other benchmark algorithms.</p>