The loss of population diversity is an important reason to cause the premature convergence in particle swarm optimization(PSO). Therefore, this paper makes the probabilistic characteristics analysis of learning parameters of PSO and proposes the relationship between the loss of population diversity and the probabilistic distribution and dependence of learning parameters. Then an adaptive learning particle swarm optimization(ALPSO) is proposed, where the modified probabilistic distribution of learning parameters is used to maintain population diversity. Meanwhile, the adaptive learning parameter with changing of evolutional state is designed to balance the global and local search abilities of particles. Experimental results show that ALPSO improves the convergence precision and effectively avoids the premature convergence.