基于K均值的带变异粒子群聚类算法  

Particle swarm optimization clustering algorithm with mutation based on K-means

作　　者：陶新民[1] 徐鹏[1] 张冬雪[1] 郝思媛[1] 

机构地区：[1]哈尔滨工程大学信息与通信工程学院,黑龙江哈尔滨150001

出　　处：《应用科技》2011年第12期25-28,共4页Applied Science and Technology

基　　金：国家自然科学基金资助项目(61074076)

摘　　要：针对K均值算法的搜索结果依赖于初始聚类中心以及粒子群算法早熟收敛的缺点,提出了一种基于K均值的带变异粒子群聚类算法.该算法通过粒子群算法来弥补K均值算法的不足,根据粒子的收敛情况判断K均值操作的时机,提高了搜索性能,并采用变异操作来跳出局部极值.分别用K均值算法、PSO-K均值算法和该算法对3种实际数据进行了聚类测试,实验结果的比较表明,该算法可以跳出局部极值,找到比其他2种算法更好的解,有更好的寻优效率并且更加稳定.To deal with the K-means algorithm＇s defects of sensitivity to the initial cluster center and the premature convergence of particle swarm optimization algorithm, a particle swarm optimization clustering algorithm with mutation based on K-means is proposed. This algorithm compensates the shortcoming of K-means algorithm by using particle swarm optimization algorithm, determines the timing of K-means operation, according to the convergence of particles, and therefore improves search performance, and jumps out of local minima by using mutation operation. The K-means algorithm,PS0- K-means algorithm and the algorithm proposed above are used to test the clustering of the three kinds of actual data. The comparison of the experimental results show that the algorithm can jump out of local minima, and it is able to find a better solution than the other two algorithms, therefore more efficient and more stable.

关 键 词：聚类分析 K均值算法 粒子群优化算法 群体智能 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]