摘要
提出了一种k-means改进算法,通过考虑样本密度、距离因素选择初始聚类中心,有效克服了经典k-means算法初始值敏感、收敛结果容易陷入局部最优解的缺点。同时引入变异系数法对样本的不同属性在聚类过程中所起的作用不同进行加权处理,全面反映了各个属性对聚类结果的影响程度。最后利用KDD Cup 1999数据集进行仿真实验,结果表明,改进算法有效地提高了入侵检测质量。
An improved K-means algorithm is presented in this paper.By considering both the sample density and the distance factor in initial cluster center selection,the sensitivity of the initial value from the traditional k-means is effectively overcame and the defect that convergence easily getting into local optimum is well solved.Meanwhile to fully reflect the clustering results of various attributes,the variation coefficient method which is greatly helpful in weighting various effects of sample-attribute is introduced in the clustering processing.And at last,the simulated experiments by data sets KDD Cup 1999 shows that the improved algorithm effectively raises the intrusion detection quality.
出处
《计算机安全》
2012年第6期2-5,共4页
Network & Computer Security
关键词
K-MEANS算法
聚类分析
变异系数法
入侵检测
K-means Algorithm； Clustering Analysis； Variation Coefficient Method； Intrusion Detection；
