CGA-ELM:A network security situation prediction model

Abstract

The combination of intelligent optimization algorithm and neural network is an efficient method for network security situation prediction, but the simple combination method can not dynamically adjust the parameters according to the existing state, which reduces the prediction accuracy. In order to get the prediction results more accurately and quickly, combined with the randomness and stability of cloud model, the global search ability and implicit parallelism of genetic algorithm and the fast learning ability of extreme learning machine, an adaptive Cloud Improved Genetic Algorithm Optimization Extreme Learning Machine (CGA-ELM) prediction model is proposed. Firstly, the crossover rate and mutation rate of genetic algorithm are improved by using the normal distribution characteristics of normal cloud. Secondly, the initial weight and deviation of extreme learning machine are optimized by the improved genetic algorithm. The simulation results show that, compared with the traditional GA-ELM, the prediction accuracy of CGA-ELM is improved by 4.9%, and the convergence speed is accelerated by 64.28%. CGA-ELM has better prediction effect and robustness. (Abstract)