<p>The influence maximization problem focuses on identifying a small subset of influential users in social networks that can ignite maximum influence spread. The existing meta-heuristic evolutionary optimizations for the problem struggle with the threat of premature convergence into local optima easily and lack flexibility. It highlights the need for solutions with high influence spread while showing satisfactory time efficiency in large-scale networks. To address such challenges, a competitive learning-driven differential evolution (CLDE) optimization is proposed. The algorithm constructs an evolutionary population by employing a random partitioning mechanism that divides it into an excellent subpopulation and a common subpopulation. The global exploration and local exploitation operations are adopted for each subpopulation to improve the solution diversity and performance. Furthermore, an adaptive probabilistic updating-driven local search strategy is incorporated to better accommodate the topological structure of networks and enhance the robustness of the algorithm. Extensive experiments conducted on six real social networks and three different types of synthetic networks demonstrate that the proposed CLDE achieves an average improvement of 6% in influence spread compared to the state-of-the-art baselines, as well as competitive results to the well-known greedy-based cost-efficient lazy forward algorithm.</p>