融合邻域扰动的简化粒子群K-均值聚类算法  

Simplified particle swarm K-means clustering algorithm for merging adjacent disturbances

作　　者：王日宏[1] 崔兴梅 周炜[1] 李祥 Wang Rihong;Cui Xingmei;Zhou Wei;Li Xiang(School of Computer Engineering,Qingdao University of Technology,Qingdao Shandong 266033,China)

机构地区：[1]青岛理工大学计算机工程学院,山东青岛266033

出　　处：《计算机应用研究》2018年第11期3232-3236,3242,共6页Application Research of Computers

基　　金：国家自然科学基金资助项目(61502262);山东省研究生教育创新计划资助项目(SDYY16023)

摘　　要：针对粒子群优化算法容易陷于局部最优,且初始聚类中心选择对K-均值算法的影响较大,提出一种融合邻域扰动的简化粒子群K-均值初始优化聚类算法(ADPSO-IKM)。根据集群度思想实现优化初始聚类中心,在粒子群算法公式中加入邻域扰动项,避免陷入局部最优,并且算法遵循自适应度优化学习策略增强全局搜索能力,进一步提高了算法精度。通过仿真测试表明,提出的ADPSO-IKM算法能加快收敛速度,可防止粒子的早熟,收敛效果好并具有较好的稳定性。In view of the particle swarm optimization algorithm was easy to be localized,and the initial clustering center had a great influence on the K-means algorithm,this paper developed a simplified particle swarm optimization for merging adjacent disturbances and K-means initial clustering algorithm(ADPSO-IKM).First,it optimized the initial clustering center based on the idea of cluster degree.Then,it added the adjacent disturbances factor in the particle swarm algorithm formula,to avoid falling into the local optimal value.Finally,the algorithm followed the adaptive optimized learning strategy,enhanced the global search capabilities,further improved the accuracy of the algorithm.The simulation experiments show that the proposed ADPSO-IKM algorithm can accelerate the convergence speed,prevents the precocity of the particles,it has good convergence effect and better stability.

关 键 词：粒子群优化算法 邻域扰动 K-均值聚类 优化初始聚类 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]