Abstract: According to biological rules of predator-prey behavior,a double-population particle swarm optimization (DPPSO) algorithm is proposed.The particles are divided into two populations,the predator population and the prey population. The particles in the predator population exclude those in prey population in a certain interval of iterations.During the course of exclusion,particles in predator population adopt the strategy of "catching the ringleader first in order to capture all his bandit followers",which means that all predator particles chase after the best position in prey population and particles in prey population try their best to escape from the nearest predator particles.In order to enhance the capacities of escaping from local optimum of the particles in prey population with stagnation state,a speed mutation method is used.The experiment results on benchmark functions show that DPPSO algorithm has the properties of fast convergence rate and strong global searching capability.