Abstract: In order to improve the searching performance of the algorithm in finding high-dimensional functions，an improved particle swarm optimization algorithm is proposed． This algorithm combines the good global search performance of AFSA ( Artificial Fish Swarm Algorithm) with the advantage of strong local search performance of PSO ( Particle Swarm Optimization) ． In initial period，AFSA was used to obtain the optimal population，and PSO was used to achieve the fine search in the later stage． In order to solve the problem of arbitrary initial population and uneven distribution，uniform initialization was used to optimize the distribution． Aimed at the poor global search direction and low efficiency of the algorithm，grouping strategy based on SFLA ( Shuffled Frog Leaping Algorithm) was adopted，and different search strategies for good individuals and other ordinary individuals in the group were used to improve the purpose and efficiency of search． Because PSO is prone to stagnation and results in low accuracy of the final result，the improved elite Gaussian learning was introduced to enhance the accuracy of the final result． The proposed algorithm is used to solve function optimization on six standard functions and compared with other algorithms，the results show the improvement is effective and superior to other algorithms．

Key words: hybrid algorithm with fish swarm-particle swarm optimization, uniform initialization, grouping strategy, elite gaussian learning