Network Security Situation Prediction Based on Sequential Extreme Learning Machine

To improve the prediction accuracy of Extreme Learning Machine (ELM), this paper establishes a sequential ELM algorithm with a forgetting factor by applying exponential weighting to new and old sample data, and obtains the optimal forgetting factor through simulation analysis. First, considering the different contribution rates of new and old data to the accuracy of the prediction model, this paper presents an ELM external weight online identification algorithm with a forgetting factor for downgrading old information. Then, based on the actual situation of network security attacks in a certain enterprise, the optimal forgetting factor for ELM external weight online learning is obtained through simulation analysis. Finally, the network security situation prediction accuracy of three ELM prediction models is compared. The simulation results show that ELM with forgetting factor has higher accuracy than traditional ELM and online sequential ELM.