Research on Data Mining of Network Security Hazards Based on Machine Learning Algorithms

Abstract

Abstract With the development and progress of science and technology, an excellent algorithm for data mining of network security hazards is sought, which can effectively discover potential dangers in the network. Based on the XGBoost machine learning algorithm, the differential evolution (DE) algorithm is used to train the XGBoost algorithm, and then an optimized DE-XGBoost algorithm is proposed. The construction of an optimal mining and evaluation model is based on this. The DE-XGBoost algorithm’s performance is assessed against cybersecurity hazards using nominal-type posture indicators when data mining cybersecurity hazards. The experimental results show that the DE-XGboost algorithm has the lowest execution time and memory usage during mining, 5min and 82MB respectively, when the number of records in the dataset is 3,500. The DE-XGboost algorithm averages a digging full rate of 92.3%, which is the highest in terms of digging full rate. The posture evaluation experiment uses the DE-XGboost model to predict the posture value that matches the real value with the maximum number of sample points, which is 10 samples. The DE-XGboost algorithm is the perfect choice for cybersecurity data mining due to its optimal performance and best mining effect.