Abstract: In order to predict the complicated network security situation reliably, a hybrid network security situation predictive model based on kernel function PSO_SVR is proposed. To solve the problem that the accuracy of prediction is low and there is no uniform to set the parameters, we construct a hybrid kernel function whose interpolation and extrapolation performance is good. The particle warm optimization (PSO) based on hybrid kernel function is introduced to search the optimization parameters. The simulation results show that the model compared with the traditional network security situation prediction method is more secure than in the prediction accuracy. The predictions are more scientific and reliable.