基于相对密度的聚类算法研究

Clustering Algorithm Research Based on Relative Density

基于相对密度的聚类算法Red的基本思想是,根据给定的半径参数求出每个点的密度,并据其对数据集中的点按照从大到小的顺序进行排序,每次均从未被聚类的点中找出密度最大的点开始聚类。聚类时,先找出一个未被聚类的密度最大的点,找到所有可达该点的点作为一类,再对剩余未被聚类的数据集中找到一个密度最大的点用同样的方法继续聚类,最后输出各个类,将不属于任何类的点作为孤立点。

In this paper, it introduces the clustering algorithm, and how to use the algorithm in details. The principle of it is to get density of each point according to radius parameter, and arrange the point clustered by data from large to small. Each time, it starts to cluster with the point with the largest density, and finds every points which shall connect with the point as one class. In the residual points, find one point with the largest density, followed by the above method, to cluster. At last, it finds each class, and puts the points not belong to every class as outlier.