CAJ | 학술논문

针对分类规则的预处理问题,提出离群属性检测分类算法.在报文分类规则属性域上计算离群属性子集,利用规则属性加权矢量计算加权距离,分析规则加权邻域的子空间离群影响因子,通过与离群因子阈值比较生成频繁匹配子集对规则进行预处理.实验结果表明,该算法能缩小后续报文的匹配范围,提高报文转发的匹配精度与速度.
침대분류규칙적예처리문제,제출리군속성검측분류산법.재보문분류규칙속성역상계산리군속성자집,이용규칙속성가권시량계산가권거리,분석규칙가권린역적자공간리군영향인자,통과여리군인자역치비교생성빈번필배자집대규칙진행예처리.실험결과표명,해산법능축소후속보문적필배범위,제고보문전발적필배정도여속도.
Aiming at the preprocessing problem for classification rule, this paper proposes outlier attribute detection classification algorithm. It accounts outlier attributes subspace on packet classification rule attribute, uses rule attribute weighted vector to calculate weighted distance, analyzes subspace outlier influence factor of rule weighted neighborhood area, and generates frequent matching subset by comparing with outlier factor threshold value. Experimental results show that this algorithm can shorten the matching rang of follow packet, enhance matching precision and speed of packet forwarding.