基于ROC曲线的逻辑回归切割值寻优方法研究及应用  

Cutting Value Optimization Method of Logistic Regression Based on ROC Curve

作　　者：邓飞 张荣稳 余靖 何俊 洪孙焱[1,2] DENG Fei;ZHANG Rongwen;YU Jing;HE Jun;HONG Sunyan(College of Information Engineering,Kunming University,Kunming 650214,China;Key Laboratory of Data Governance and Intelligent Decision in Universities of Yunnan,Kunming University,Kunming 650214,China)

机构地区：[1]昆明学院信息工程学院,云南昆明650214 [2]云南省高校数据治理与智能决策重点实验室,昆明学院,云南昆明650214

出　　处：《昆明理工大学学报（自然科学版）》2023年第1期53-57,共5页Journal of Kunming University of Science and Technology(Natural Science)

基　　金：国家自然科学基金项目(62066023);云南省地方本科高校基础研究联合专项-面上项目(202101BA070001-149);云南省教育厅科学研究基金项目(2022J0634)。

摘　　要：针对传统切割值准确率预测较低的问题,提出了一种逻辑回归切割值寻优的方法.首先针对传统逻辑回归模型中切割值大多默认选取0.5,所造成的预测准确率较低的问题,利用逻辑回归模型来预测网络攻击,结合ROC曲线,将不同的切割值的准确率进行对比,提高预测的准确率.最后通过网络入侵检测实验,证明了该方法寻找到的切割值所对应的准确率,比传统的切割值的准确率有大幅提高.A logistic regression cut value optimization method was proposed to solve the problem of low accuracy of traditional cut value prediction.In view of the low prediction accuracy caused by the default selection of 0.5 in the traditional logistic regression model,the logistic regression model is used to predict network attacks based on the ROC curve,and the accuracy of different cut values is compared to improve the prediction accuracy.The network intrusion detection experiment proves that the accuracy of the cutting value found by this method is significantly higher than that of the traditional cutting value,which verifies the effectiveness and scientificity of the method.

关 键 词：网络入侵 逻辑回归 ROC曲线 最优切割值 CIC-IDS 2017 

分 类 号：TP393.08[自动化与计算机技术—计算机应用技术]