Abstract: Influenced by a variety of complicated factors, the network security situation has many characteristics, such as highly nonlinear, time-varying, and mutant. It is difficult to predict accurately with a single prediction method. In response to this shortage, a new combined prediction model for network security situation was proposed based on cooperation policy theory. The network security situation was predicted respectively by using the Elman neural network model, GM(1,1)model, support vector machine (SVM) mode. The Shapley value method of cooperative games was applied to determine the weight of each single prediction model, and the prediction results were weighted calculated to get the final combined prediction results of network security situation. The actual network security data were used for simulation testing. The simulation results show that combination prediction model can effectively improve the network security situation prediction accuracy.

Key words: network security situation, Elman neural networks, GM(1,1), support vector machine(SVM), combination prediction