<p>Swarm intelligence optimization has been widely utilized in many fields with the growth of scientific engineering and mathematics, and clustering is a data mining problem that requires settling theoretically. Since the emotional preference and migration behavior clustering (EPMC) model has efficiently solved the clustering tasks, it also suffers from premature convergence and the population diversity problem. Besides, the solution searching range of EPMC is too broad, and the evolution process is entirely random. Therefore, in order to enhance the optimization ability of the EPMC algorithm, we incorporate several evolution strategies such as the non-linear position weight, the directional position increment, and the novel non-linear cosine adaptive crossover and mutation method into the emotional preference and migration behavior clustering (EPMC) model. As a result, an emotional preference migration clustering model based on position increment nonlinear cosine adaptive crossover and mutation strategy (DPNG-EPMC) was proposed. The DPNG-EPMC can further prevent the algorithm from prematurely falling into local optimality similar to gradient descent, improve the diversity of the population, and enrich the family of the clustering methods. The proposed DPNG-EPMC is compared with the other nine clustering algorithms through numerous experiments, and the results of several criteria show the effectiveness of our proposed ensemble algorithm. Besides, theoretical analysis has verified the convergence of the DPNG-EPMC algorithm.</p>