<p>In success history-based differential evolution, a common strategy is to select individuals with superior fitness values to guide the population in exploring promising regions. However, some recent studies show that the linear combination vector has a greater chance of converging towards the optimal solution than the current best vector, from both theoretical and experimental perspectives. Since the linear combination vector can better approach the optimal solution when individuals are clustered, it is necessary to use a clustering method to assess their aggregation. Based on this idea, this paper proposes a novel method named Nearest Better Clustering-based Linear Combination (NBCL) to generate best vectors. The method uses nearest better clustering to divide the population into several clusters; the individuals within each cluster are then linearly combined according to the superiority and inferiority of their fitness values to generate best vectors. Several representative and high-performance success history-based differential evolution algorithms were tested using this method on CEC2017 and CEC2022 benchmark functions. The experimental results show that the NBCL method effectively improves the convergence speed of the algorithm while ensuring accuracy. A diversity analysis was also conducted, indicating that the NBCL method effectively maintains population diversity. Meanwhile, parameter configuration experiments validated the effectiveness of the parameter settings.</p>