Abstract: This study presents a modified Shuffled Frog Leaping Algorithm (SFLA) which uses a new accelerated method and Gaussian mutation to control its search behavior. Aims to search the solution space more flexible, random disturbance accelerated strategy provides a dynamic expanded neighbor structure. The policy of probabilistic selection is introduced to save the function evaluation times. A Gaussian mutation is also evolved to generate new frog randomly without sacrificing the diversity of the algorithm. Experiments with a wide range of benchmark functions demonstrate good performance of the proposed algorithm when compared with the classic SFLA and other recent variants of SFLA in terms of global optimality, solution accuracy, algorithm reliability and convergence speed.