信息安全评估日志数据的一种混合聚类算法  

A Hybrid Clustering Algorithm for Information Security Evaluation Log Data

作　　者：陈庆枝[1] 陈国龙[2] 郭文忠[2] 陈仕涛[2] 

机构地区：[1]莆田学院电子信息工程学系,福建莆田351100 [2]福州大学数学与计算机科学学院,福州350108

出　　处：《重庆工学院学报（自然科学版）》2009年第10期77-82,118,共7页Journal of Chongqing Institute of Technology

基　　金：国家自然科学基金资助项目(60673161);福建省自然科学基金重点项目(A0820002)

摘　　要：首先引入能够处理混合型数据的K-prototypes聚类算法,在此基础上构造了一种基于粒子群优化算法和K-prototypes方法的混合聚类算法.利用粒子群优化算法良好的全局搜索能力,克服K-prototypes容易陷入局部最优值的不足.实验结果表明,该算法能够避免陷入局部最优值,具有较好的全局收敛性,并且提高了聚类的正确率和算法的稳定性.There are many categorical and numericalattributes in source data of information security evaluation. When dealing with mixed data, the traditional cluster algorithms tend to converge to a local optimum or can only process numerical data. In order to solve these problems, this paper employs the K-prototypes cluster algorithm to deal with mixed numerical data, and a hybrid cluster algorithm is designed based on particle swarm optimization algorithm and K-prototypes algorithm, combing the K-prototypes algorithm＇ s ability on local search and fast convergence and the particle swarm optimization algorithm＇ s ability on global search. Experimental results show that the hybrid algorithm has better performance, strong robustness and fast convergence speed.

关 键 词：聚类 信息安全 安全评估 粒子群优化 K—prototypes 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]