最小生成树聚类方法研究

Research of the Minimum Spanning Tree Clustering Algorithm

由聚类所生成的簇是一组数据对象的集合,在同一个类中的对象之间具有较高的相似度,而不同类中的对象差别较大.图的最小生成树具有最优子结构性质,删除最小生成树的最大边后的两颗子树依然分别是两个子图的最小生成树,因此可由生成图的最小生成树获得聚类.此方法适用于所有欧氏空间数据的聚类.

The cluster from clustering is an aggregate of some data object. There is the higher likeness degree in the same object, and there is difference degree in the different cluster greatly. The minimum spanning tree of the graph has the superior sub - structure property, The sub trees that delete the biggest side respectively are two minimum spanning trees still, So we can be been clustering by the minimum spanning tree of the complete graph. This method is applicable to all Euclidean spatial data to clustering analysis.