基于自适应遗传算法的混合特征选择方法  

HYBRID FEATURE SELECTION METHOD BASED ON ADAPTIVE GENETIC ALGORITHM

作　　者：裴作飞 李兆玉[1] 王云锋 姚立霜 Pei Zuofei;Li Zhaoyu;Wang Yunfeng;Yao Lishuang(School of Communication and Information Engineering,Chongqing University of Posts and Telecommunications,Chongqing 400065,China)

机构地区：[1]重庆邮电大学通信与信息工程学院,重庆400065

出　　处：《计算机应用与软件》2020年第8期256-259,306,共5页Computer Applications and Software

基　　金：长江学者和创新团队发展计划项目(IRT_16R72)。

摘　　要：高维数据含有大量冗余和噪音特征影响检测效果,维度过高使得系统训练时间长、实时性差。采用卡方(Chi Square)过滤算法,删除冗余和相关性低的特征;采用LightGBM算法作为封装方法组成混合特征选择算法,通过自适应遗传算法进行搜索获取最优特征子集。在入侵检测KDDCUP99数据集上进行3种算法的对比验证。实验结果表明,该方法具有较好的检测效果和特征约减能力。There are a lot of redundancy and noise characteristics in high-dimensional data that affect the detection effect.The high dimension makes the system training time long,and the real-time performance is poor.The Chi Square filtering algorithm was used to remove the features of redundancy and low correlation.The LightGBM algorithm was used as the wrapper method to form a hybrid feature selection algorithm,and the adaptive genetic algorithm was used to search to obtain the optimal feature subset.Finally,we compared and verified three kinds of algorithms on the intrusion detection KDDCUP99 dataset.The experimental results show that our method has better detection effect and feature reduction ability.

关 键 词：遗传算法 混合特征选择 入侵检测 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]