K-means聚类算法优化方法的研究

针对K-means算法全局搜索能力的不足,提出基于改进PSO的优化K-means聚类算法(IPSO-KM),该算法克服了K-means聚类算法对初始聚类中心选择敏感问题,能够获得全局最优的聚类划分.同时,提出一种基于信息增益比例的属性加权的实体之间距离计算方法,使用属性加权距离计算方法进行聚类划分时,无论是球形数据还是椭球形数据都能够获得较好的聚类划分结果.仿真实验采用KDD-cup 99的测试数据,实验结果表明本文提出的算法不但能检测到多种已知的网络入侵行为,而且能够检测到许多未知的网络入侵行为,同时保持较高的网络入侵的检测率和较低入侵的误报率.

Research on Optimization Method for K-means Clustering Algorithm

Aimed at the lack of global search capability of K-means algorithm,optimized K-means clustering algorithm based on improved particle swarm optimization(IPSO-KM) is presented in this paper,which can overcome the problem of initial clustering center selection sensitivity of K-means and can obtain global optimized clustering partition.At the same time,an attribute-weighting distance computation method based on information gain ratio is presented,the better clustering partition can also be obtained for whether spherical data or ellipsoidal data.Simulation experiment is implemented over data set KDD-cup 99,the experimental result showed that the algorithm presented in this paper not only detected many known network intrusion behaviors,but also detected many unknown network intrusion behaviors,at the same time keeping the higher detection rate and lower false acceptance rate.