A Modified Mountain Clustering Algorithm based on Hill Valley Function
A Modified Mountain Clustering Algorithm based on Hill Valley Function
复制标题
DOI：
10.4304/jnw.6.6.916-922
复制
发表时间：
2011-01
期刊：
影响因子：
--
通讯作者：
Junnian Wang;Deshun Liu;Chao Liu
中科院分区：
文献类型：
--
作者：
Junnian Wang;Deshun Liu;Chao Liu
提出了一种基于山谷函数的改进山地聚类算法。首先，在数据空间上构建山函数，通过相关自比较方法估计参数，计算数据库的山函数值。其次，引入山谷函数对分布在各个峰上的数据进行划分。如果两个基准的山谷函数值等于0，则说明这两个基准在同一座山上，属于同一簇，否则不是。最后选择山函数值最大的簇中的数据作为该簇的簇中心。对四个数据库的测试表明，所提出的聚类算法能够对每个簇中的数据数进行分类，准确地找到所有的聚类中心，并且不需要与数据库相关的先验参数和停止准则。
A modified mountain clustering algorithm based on the hill valley function is proposed. Firstly, the mountain function is constructed on the data space, with estimating the parameter by a correlation self-comparison method, and database’s mountain function values are computed. Secondly, the hill valley function is introduced to partition the data distributed on each peak. If the hill valley function’ value of two datum equal to 0, it means these two datum are on the same mountain and belong to the same cluster, otherwise they are not. Finally, the data in a cluster with maximum mountain function value is selected as the cluster centre of this cluster. The testing of four databases indicate that the proposed clustering algorithm can categorise the data numbers in each cluster and find all the cluster centres exactly, and no need priori parameters and stopping criterion correlating to the database.