基于群优化算法的聚类分析法

发布时间：2018-01-31 04:20

本文关键词： 磷虾群优化算法粒子群优化算法聚集度聚类分析　出处：《北方民族大学》2017年硕士论文　论文类型：学位论文

【摘要】：聚类分析法是统计学中的一种重要分析技术,群智能算法由于其高效的优化处理能力而越来越受到人们的重视.本文介绍了聚类分析以及群优化算法的相关知识,对标准磷虾群优化和粒子群优化算法进行了分析和研究,提出了基于聚集度改进的异化磷虾群算法(CSKH)和基于聚集度的改进粒子群优化算法(CSPSO).最后将这两种改进算法整合到聚类分析算法中,并通过真实数据集测试了整合算法的性能.具体内容如下:首先,针对标准磷虾群算法存在着不易跳出局部寻优、搜索精度低等问题,提出了一种基于聚集度的异化磷虾群算法.本算法根据种群多样性指标聚集度的变化,通过在两个相反位置移动方向的选择策略来增加磷虾进化多样性,同时引入了随机数策略来模拟磷虾的外部扰动,从而取代原磷虾群算法中的随机扩散运动的影响.算法还引入平均距离指标来增加局部搜索的变异概率,同时将背向最优位置的速度方向作为搜索变异方向,从而扩大了群体的搜索空间,保证算法的全局搜索能力.其次,针对粒子群算法极易出现早熟收敛的问题,在重新定义相似度的基础上构建了聚集度概念,据此来描述种群的多样性程度,并通过设定粒子群自适应阈值的变化来调整粒子搜索空间,同时根据聚集度的大小对粒子重新赋值,增加种群的多样性,从而使得算法更易于跳出局部最优.最后,将上述两种改进算法整合到聚类方法之中,通过改进算法的收敛能力来指引聚类的进行方向,并通过对两组真实数据集的检测验证了整合算法的性能.
[Abstract]:Cluster analysis is an important technique in statistics. Swarm intelligence algorithm has attracted more and more attention because of its high efficiency. In this paper, clustering analysis and knowledge of swarm optimization algorithm are introduced. The standard krill swarm optimization and particle swarm optimization algorithm were analyzed and studied. In this paper, an improved clustering algorithm named CSKH and an improved particle swarm optimization algorithm based on aggregation are proposed. Finally, these two improved algorithms are integrated into the clustering analysis algorithm. The performance of the integration algorithm is tested by real data set. The main contents are as follows: first of all, the standard krill swarm algorithm is difficult to jump out of the local optimization, the search accuracy is low and so on. An algorithm based on aggregation degree is proposed to increase the evolutionary diversity of krill, which is based on the change of population diversity index and the selection of moving direction in two opposite positions. At the same time, the random number strategy is introduced to simulate the external disturbance of krill, so as to replace the random diffusion motion in the original krill swarm algorithm. The average distance index is also introduced to increase the mutation probability of local search. At the same time, the velocity direction of the optimal position is taken as the direction of search mutation, which expands the search space of the population and ensures the global search ability of the algorithm. Secondly, the particle swarm optimization algorithm is prone to premature convergence. Based on the redefinition of similarity, the concept of aggregation degree is constructed to describe the diversity of population, and the particle search space is adjusted by setting the adaptive threshold of particle swarm. At the same time according to the size of the aggregation of particles re-assigned to increase the diversity of the population so that the algorithm is easier to jump out of the local optimal. Finally the above two improved algorithms are integrated into the clustering method. By improving the convergence ability of the algorithm to guide the direction of clustering, and through the detection of two groups of real data sets to verify the performance of the integrated algorithm.
【学位授予单位】：北方民族大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP18;TP311.13

【参考文献】