摘要
为了克服经典K-Means算法随机选择初始数据中心而易陷入局部最优解和聚类结果的不确定性问题,提出一种基于粒子群和K-Means算法的改进聚类算法以实现移动用户分类。首先,定义数据对象密度并采用改进的普里姆算法初始化聚类中心,然后,将此聚类中心用于初始化粒子位置,采用混沌粒子群算法寻优获得最优解作为最终的聚类中心,最后,采用经典K-Means算法根据最终聚类中心进行聚类。仿真实验表明文中方法能正确地实现移动用户分类,并具有较强的全局寻优能力和较快的收敛速度,弥补了经典K-Means方法的不足,具有较强的现实意义。
In order to conquer the clustering result uncertainty and easily obtaining the local optimum solution of random choosing initial data center in K-Means algorism, a improved algorism based on （Particle swarm optimization algorism, PSO） and K-Means used to realize the classification of mobile users is proposed. Firstly, the data object density is defined to improve prim algorism, and then the improved prim algorism was used to initialize the clustering center, then the clustering center is used to initialize the position of the particles, and the chaos-PSO algorism was used to get the global optimum solu- tion, finally, the classic K-Means algorism was operated to cluster according to the final optimum clustering center. The simulation experiment shows the method in this paper can realize the classification for mobile users, and has the strong glob- al optimizing ability and convergence rate, making up the defects of classic K-Means method. It is proved to have the strong practical significance.
出处
《计算技术与自动化》
2013年第4期57-60,共4页
Computing Technology and Automation
基金
江苏省教育科学"十二五"规划2011年度课题项目(D/2011/03/006)
关键词
粒子群
K均值
分类
聚类
particle, K-Means
classification
clustering
