• 中国计算机学会会刊
  • 中国科技核心期刊
  • 中文核心期刊

计算机工程与科学

• • 上一篇    下一篇

基于机器学习的KDD-CUP99网络入侵检测数据集的分析

余华鸿,周凤艳,陈毛毛   

  1. (陆军防化学院基础部,北京 102205)
  • 收稿日期:2019-08-06 修回日期:2019-10-17 出版日期:2019-12-25 发布日期:2019-12-25

Analysis of KDD-CUP99 network intrusion
detection data set based on machine learning

YU Hua-hong,ZHOU Feng-yan,CHEN Mao-mao   

  1. (Ministry of Basic,the Army Institute of Chemical Defense,Beijing 102205,China)
  • Received:2019-08-06 Revised:2019-10-17 Online:2019-12-25 Published:2019-12-25

摘要:

使用Python编程,采用朴素贝叶斯分类器、Softmax回归和决策树回归3种有监督学习算法,对KDD-CUP99网络入侵监测数据集进行训练,并分析结果。首先通过3种分类器库的函数,对KDD-CUP99数据集进行分析预测;然后通过增量式训练方法探究3种分类器对训练数据量的依赖程度;最后通过特征筛选探究3种分类器算法受样本特征数量的影响程度。

关键词: 机器学习, 模型训练, 分析预测, 有监督学习

Abstract:

This paper adopts machine learning methods to train the kdd-cup99 network intrusion detection data set and analyze the results. Python programming is used to implement three supervised learning methods such as “naive Bayes classifier”, “Softmax regression” and “decision tree regression”. Firstly, the three classifier library functions are used to analyze and predict the KDD-CUP99 data set. Then, the incremental training method is used to explore the dependence of three classifiers on the amount of training data. Finally, feature screening is adopted to explore the influence of the number of sample features on the three classifier algorithms.
 

Key words: machine learning, model training, analysis and prediction, supervised learning