Aiming at improving the atifical fish swarm algorithm which is easy to fall into local optimal result in late iterative search,an improved fish swarm algorithm was proposed based on knockout rules——elite fish swarm algorithm(LOAFSA).LOAFSA involves a knockout system based on pareto principle and a population dynamics management strategy based on Gauss optimization.The experiments show that,for the extremal problems of multi-variable function,it has a good global search ability.The algorithm was applied to solve the mathematical optimization model of RAN architecture of the wireless access network to find out the optimal set and to get the maximum capacity.