Abstract: SVR (support vector regression) has a great advantage in solving nonlinear regression problems. In the process of SVRs predictions, the most important step is the choice of parameters. The result will be very different because of the change of parameters. The common method is to use GA(genetic algorithm) and PSO(particle swarm algorithm) for parameter selection, however, the limitations of these two algorithms in solving the problem of multi-modal can easily lead to low efficiency and the accuracy is not high. Cuckoo search algorithm introduces a Lvy flight mechanism that can effectively escape from local optimal solution. The algorithm converges fast, and the result is not sensitive to the parameters of the algorithm itself. The cuckoo search algorithm is applied to the parameter selection of SVR in this paper. The experimental results of network traffic prediction and the wine quality prediction show that the cuckoo search algorithm is faster and better compared with GA, PSO algorithms.