Parameter optimization of support vector machines has always been an important research direction. The quality of the parameters determines the classification accuracy and generalization ability of the support vector machine (SVM) to a great extent. Given that using the artificial fish swarm algorithm for optimizing the parameters of the SVM tends to hover near the optimal solution in the late stage, we put forward an improved artificial fish swarm algorithm. It uses velocity parameters instead of the artificial fish step and obtains optimal target and the optimal parameters of the SVM. Simulation results show that the proposed method has the advantages of fast convergence speed, high numerical accuracy and low dependence on initial values, as well as better performance and higher classification accuracy in SVM parameter optimization. It is an effective parameter optimization method.