摘要
特征选择是模式识别领域的一个重要的研究方向,它可以提高分类的效率与效果。本文将递归特征排除算法与SVM决策树结合起来运用于特征选择,首先利用递归特征排除算法对所选择的特征进行初排序,然后依次将特征送入SVM决策树中进行优化评估,对数据中起显著作用的特征进行筛选,除去冗余和次要特征,得到特征子集。最后,通过对Linux主机和相关网络的27个入侵特征数据进行特征选择实验,实验结果表明,特征个数降至21个,而测试精度仍然能达到94%,从而证明本文所提出的递归和SVM相结合的方法是解决特征选择问题的一种有效方法。
Feature selection is an important field of pattern recognition research,it can improve the efficiency and effectiveness of classification.This is paper uniting recursive feature exclude algorithm with SVM decision tree applied to feature selection,first use recursive feature exclude algorithm to sort for selection characteristics,and then it was send into the SVM decision tree to optimize assessment,remove redundancy and secondary features,aim to filtering significant feature subset.Finally,through Linux hosts and related 27 network intrusion feature the test data conduct feature selection experiment,The result show that the feature number can be reduced to 21,while the test accuracy still reach 94%.So the proposed method of uniting recursive with SVM are effective for feature selection.
出处
《电子测试》
2010年第9期26-29,92,共5页
Electronic Test
关键词
特征选择
递归
SVM
入侵检测
Feature selection
recursive
SVM
intrusion detection
