【摘要】 针对现有集成入侵检测算法的多样性不足或基分类器强度不足问题,提出基于SPCA的增强型旋转森林算法(SPCA-ERoF)。通过引入旋转作用增加集成多样性,为提升旋转数据可分性,分析PCA算法,提出SPCA算法;利用随机森林作为旋转森林的基分类器,解决基分类器强度不足,进一步提升集成多样性。实验结果表明,SPCA相比PCA能在一定程度上提升旋转数据的可分性;SPCA-ERoF在KDD99入侵数据集上取得了良好的检测结果,各类样本的综合性能指标F1_Score均能达到90%以上,有效提升了入侵检测系统的性能。更多还原

【Abstract】 Aiming at the problems of insufficient diversity or insufficient strength of the base classifier of existing ensemble intrusion detection algorithms,an enhanced rotation forest algorithm(SPCA-ERoF)based on SPCA was proposed.The rotation effect was introduced to increase the diversity of integration,at the same time,to improve the separability of the rotation data,the PCA algorithm was analyzed and the SPCA algorithm was proposed.The random forest was used as the base classifier of the rotation forest to solve the lack of strength of the base classifier and further improve the integration diversity.Experimental results show that SPCA can improve the separability of the rotation data to a certain extent compared with PCA.SPCA-ERoF achieves good detection results on the KDD99 intrusion dataset,and the comprehensive performance indexes F1_Score of various samples can reach more than 90%,which effectively improving the performance of the intrusion detection system.更多还原