Usability is crucial for mobile app success. Usability evaluation and correction of Mobile User Interface (MUI) is fundamental phase to discover the usability smells and improve its quality. However, it continues to be a hard, expensive, and generally it is ignored. Several existing works depend on a subjective assessment of usability defects which relies on users’ feedback, that leads to time-consuming, human-centric, and error-prone task. Hence, recent works proposed tools for collection and analysis log user interaction when using the mobile application to understand user behavior and identify usability smells. But, these tools treat the defects with the same importance and rarely propose recommendations to help evaluator to improve their MUIs. It still also needed to identify concrete usability defects and fixed riskiest usability smells. In addition, most existing methods proposed to assess the usability of once the applications has been developed, which led to a lot of reworking by the designer because these changes can affect the development phases.In this paper, we propose an automated approach based on correction rules which combine literature evaluation metrics with usability smells to define recommended refactoring of MUI.To this end, we consider this refactoring recommendation as a multi-objective optimization problem. The purpose of this optimization is to maximize the number of refactoring per individual while minimizing the number of features per rule, the number of rules by solution, the misclassification rate, the information gain and improving the MUI quality in terms of guidance and cohesion coverage and minimizing the complexity of MUI. Moreover, we leaded a comparative study of four evolutionary algorithms to solve this optimization problem. The results prove the efficiency of the IBEA in terms of generating the most accurate refactoring rules with an average of 0.92 of AUC.This paper applied an evolutionary algorithm to recommend the best refactoring operations to enhance the quality of MUI. The results are promising. Our search-based algorithm successfully searches the best trade-off between eight objectives and recommending the most accurate refactoring rules compared with other studied algorithms.