• 主题
高级搜索  >
历史搜索 清空历史
首页> 外文会议>International Symposium on Computer and Information Sciences(ISCIS 2005); 20051026-28; Istanbul(TR) >Effective Clustering by Iterative Approach
【24h】

Effective Clustering by Iterative Approach

机译：迭代法有效聚类

获取原文
获取原文并翻译 | 示例
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

In this study, we present multi-objective genetic algorithm based iterative clustering approach. Two objectives are employed in the process: minimizing the within cluster similarity and maximizing the difference between the clusters: inter-cluster distance (average linkage, centroid linkage, complete linkage and average to centroid linkage) versus intra-cluster distance (total within cluster variation). The proposed approach is iterative in the sense that it basically tries possible partitioning of the dataset for the given range of clusters one by one; the result of the previous partitioning n favors that of the current solution n+1. In order to achieve this, we identified a global k-means operator and we do "what if analysis in the aspect of the objectives to see the better initialization in case the number of clusters is increased by one. After evaluating all, a feedback mechanism is supplied at the back-end to analyze the partitioning results with different indices. The entire system has been tested with a real world dataset: glass. The reported results demonstrate the applicability and effectiveness of the proposed approach.
机译：在这项研究中，我们提出了基于多目标遗传算法的迭代聚类方法。在此过程中采用了两个目标：最小化集群内相似度并最大化集群之间的差异：集群间距离（平均链接，质心链接，完整链接以及平均到质心链接）与集群内距离（集群内总变化） ）。从某种意义上说，所提出的方法是迭代的，它基本上是针对给定范围的聚类一一尝试尝试对数据集进行分区；先前分区n的结果优先于当前解n + 1的结果。为了实现这一目标，我们确定了一个全局k均值算子，并且“如果在目标方面进行分析，以防在簇数增加一个的情况下更好的初始化该怎么办。在评估了所有结果之后，建立了反馈机制后端提供了用于分析具有不同索引的分区结果的系统，并已通过真实数据集（玻璃）对整个系统进行了测试，报告的结果证明了该方法的适用性和有效性。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱：kefu@zhangqiaokeyan.com

京公网安备：11010802029741号 ICP备案号：京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有

  • 客服微信

  • 服务号