In order to more accurately master the laws of network security situation regular and prevent some network security threats, in view of the problem of parameter selection of Support Vector Machines, plus, the gravitational search algorithm (GSA) is characterized by few parameters needed and having great ability in global optimization, a network security situation prediction model (GSASVM) is proposed for GSA optimization SVM parameters. First, the parameters of SVM are treated as objects in space, and mean square error (MSE) of predicted value and actual value of SVM under this parameter is used as the objective optimization function, then GSA can find the optimal parameters of the SVM by simulating the law of gravitation, and finding the optimum parameter eventually. Finally, a network security situation prediction model is established according to the optimal parameters. Using DARPA 1999 data set provided by MIT Lincoln Laboratory in MATLAB platform, the simulation results show that GSA-SVM improves the accuracy of network security situation prediction and accelerates prediction of network security situation relative to other prediction algorithms. This provides a new way to solve problem of network security situation prediction.