异常流量检测中的特征选择

针对异常流量检测领域的高维数据降维问题，提出了一种基于信息熵理论的特征选择算法。首先计算特征的重要系数，删除重要系数小于一定阈值的特征，得到重要特征集。然后，计算特征间的冗余系数，删除冗余特征，得到精简的特征集。最后，用ID3算法对精简的特征集进行了验证，结果表明这种特征选择算法是有效的。

Feature selection algorithm toward abnormal traffic detection

Focusing on the problem of dimension reduction in abnormal network traffic detection in LAN,a novel algorithm for feature selection is proposed based on information entropy.Firstly,the features whose importance coefficient is less than a threshold are deleted,and the important features set is obtained.Secondly,the redundancy features are deleted through computation of redundancy coefficient between two features.At last,the ID3 algorithm is used to detect abnormal traffic with the reduction of feature set.And the results show that the proposed algorithm to select feature is effictive.