<p>The most basic operation in developing a movie recommender system is clustering. Unfortunately, clustering is one of the NP-hard problems. Therefore, finding the optimal number of clusters and convergence speed is one of the most important concerns in recommender system development. One of the well-known methods for solving NP-hard problems is metaheuristics. However, the main problem of metaheuristic algorithms is the dependence of solutions on the location of initial centers or getting stuck in local optima. It is due to the lack of a trade-off between exploration and exploitation, two fundamental search behaviors. One of the efficient methods to balance exploration and exploitation is the use of Differential Evolution (DE). This trade-off is mainly affected by the changes applied to the mutation/crossover operators and the corresponding parameters. An important advantage of the DE method is that it does not waste time checking weak areas. Therefore, it can also increase the convergence rate by using the optimal number of clusters. Our research area in this paper is both "rank prediction" and "item recommendation" using a combined method. In this research, the combination of DE and Bees Algorithm (BA) is used to establish the trade-off between exploration and exploitation during clustering. The evaluation results show that the convergence rate of clustering with the proposed combinatorial approach is much faster than the baseline methods. In addition, it has a less intra-cluster error during training and testing steps. One of the significant implications of our research is to find the best hyperparameter settings on the MovieLens dataset.</p>