Abstract: A high efficient evolutionary algorithm for function optimization is proposed, the algorithm uses hybrid selection strategy which combines tournament selection with roulette selection, classific mutations are employed in different individuals: pattern search aims the best individual so far, cauchy mutation aims to the individuals whose order based on fitness is one third of the population while ordinary mutation is employed in other individuals. The experiments of 15 benchmark problem show that the new algorithm can not only converge fastly, but also obtain the solution with high precision even if not the best.