Mobile network risk user recognition based on ensemble learning

Abstract

With the rapid development of mobile network, there are also some hidden dangers of network security. In order to study the problem of risk user identification in mobile network, based on user desensitization data provided by China Unicom, this paper establishes a risk user identification model using four ensemble learning methods: random forest, Adaboost, Xgboost and Lightgbm, and compares the effect of the model with the accuracy, AUC and F1 scores as evaluation indicators. The results show that the ensemble learning method can effectively identify mobile network risk users. Adaboost, Xgboost and Lightgbm based on boosting algorithm have higher accuracy and stronger generalization ability than random forest based on bagging algorithm. Therefore, mobile network companies can prevent the risk of mobile network by establishing relevant ensemble learning model, and then provide a healthy mobile network environment for the public.