Nowadays, it is very difficult to distribute loads on edge servers in Mobile Edge Computing (MEC) due to the large amount of statistics and the large number of user requests. In order to handle these issues, it is necessary to propose an approach that can combine features of optimization techniques with machine learning (ML) techniques. Optimization techniques are used to handle load-balancing tasks such as managed resource availability, task scheduling, and network overhead. Furthermore, ML plays a crucial role in predicting load and user behavior based on nodes’ past histories. In this paper, we apply the reinforcement learning approach to optimal decision-making and efficiently utilize resources. We analyze the recital of the projected approach on a real dataset that we collected from an open-source platform. The results show that in the proposed approach, Fraction in Service Level Agreement (SLA) and average response time is low compared to other techniques.