To address the shortcomings of the marine predators algorithm (MPA) in solving complex problems, such as low optimization accuracy and easily falling into local optimization, this paper proposes an improved marine predators algorithm based on group learning (GLMPA). An opposition-based learning method is adopted to enhance the quality of the initial solutions. Then, a group learning strategy is used to diversify the population. Two subgroups are produced by fitness evaluation and employing different updating mechanisms. In addition, a new position-updating rule is used to help the proposed algorithm escape from the local optima in the later stage of iteration. Finally, six test functions are utilized to test the GLMPA, and the simulation results verify the effectiveness of the proposed algorithm when compared with other famous algorithms.