The Remora Optimization Algorithm (ROA) is a novel proposed metaheuristic optimization algorithms based on the swarm intelligent behavior. In ROA, the search agent simulates the movement habits of other marine creatures like humpback whale and sailfish by attached itself to the host. Thereby, ROA can contain different position updated model. However, it still leads to the local optima stagnation and the low convergence rate. Hence, we proposed a novel improved ROA with Ensemble Mutation Strategy (CMSROA) to increase the population diversity for boosting the searching efficiency. Moreover, CEC2020 test suite is employed to conduct a comparison experiment for comparing the performance between CMSROA and other outstanding metaheuristics. At the same time, Wilcoxon's Rank-Sum test is employed to verify the validation of our statistical experiment results. Finally, a classical welded beam design problem is introduced to test the capability of solving real world engineering problems. Generally, we can demonstrate the competitive performance of CMSROA based on the experiment analysis.