基于改进差分进化的K均值聚类算法在入侵检测中的研究  

Research on K Mean Clustering Algorithm Based on Improved DEAlgorithm in Intrusion Detection

作　　者：王广[1,2] 张晓明[1] 徐日华[1,2] 

机构地区：[1]北京化工大学,北京100029 [2]北京石油化工学院,北京102617

出　　处：《北京石油化工学院学报》2016年第3期28-31,共4页Journal of Beijing Institute of Petrochemical Technology

摘　　要：针对K均值算法对初始聚类中心需要人为设定、对聚类中心敏感并且极易陷入局部最优的缺陷,用改进的DE算法对K均值算法进行优化。在DE算法中,采用动态交叉参数CR与缩放参数F,有效地平衡了DE算法的全局寻优能力与收敛速度二者的矛盾;利用混沌的随机性完成DE算法的种群初始化,利用其遍历性在DE进化后期的最优解附近进行混沌搜索,有效地提高了DE算法的全局寻优能力。最后,使用KDD99数据集对CDE-K均值算法进行验证,实验结果表明,CDE-K均值算法具有较好的聚类能力,在检验效果方面表现优秀。While using the K algorithm, the mean value of K-means to the preliminary cluster center needs tobe set artificially, and the cluster center is sensitive, whicheasily leadsto thedefect ofpartial optimum. The researchers adopt the improved DE algorithm to optimize K algorithm. Using dynamic cross parameters of CR and the scaling parameter F, can effectively balance the general searching capacity and constriction speed. The chaos randomly completethe population initialization of DE algorithm and reuse characteristics of the ergodicity of the near optimal DEin- dividuals in chaotic search, which can effectively improve the DEalgorithm with global searching ability. Finally, the KDD99 data set is used to verify the optimization of the K algorithm. The results indicate that the CDE-KM has better clustering capacity and good performance in terms of test results.

关 键 词：入侵检测 K均值算法 混沌搜索 DE算法 

分 类 号：TP37[自动化与计算机技术—计算机系统结构]