The final objective of the Recommender Systems (RSs) is to offer recommendations to the user that are close to his/her taste. When the user enters the system, the most similar data cluster to the user’s taste can be selected, and by creating a neighborhood of the users similar to him/her within the selected cluster, the proposal generation can be followed. Determining the appropriate number of neighbors of the user can lead to increased accuracy of the recommendations made. Due to the enormous size of the datasets, this requires more time. This study aimed to propose recommendations with the highest accuracy and in the shortest possible time through finding the best number of neighbors for the user applying the NSGA-II Multi-Objective Evolutionary Algorithm (MOEA). Here two objects of accuracy and time of recommendations are in a multi-objective state, thus a balance should be created between the two conflict objects. The simulation results on the MovieLens 100K, MovieLens 1M, Netflix and FilmTrust standard datasets indicated that the proposed MOEA was capable of providing recommendations with greater accuracy and at the proper time, hence it could improve the Mean Absolute Error (MAE), Root Mean Square Error (RMSE), Coverage, Precision, Recall and Fmeasure criteria.