多维数据的聚类结果可视化技术综述

在很多情况下,人们不仅需要聚类算法给出类标,还需要掌握聚类结构和数据分布情况.为满足后一项需求,出现了许多聚类结果的可视化(简称聚类可视化)技术,以图形的方式将多维数据和其聚类结果显示在二维或三维空间.从直接在二维或三维空间显示数据及其聚类结果、数据经降维(或映射)后显示以及其它显示方式3种角度综述了常用的30多种聚类可视化方法,并对各种方法的优缺点和适用性进行了分析和讨论.

Study of Visualization Methods of Clustering Results for Multi-dimensional Data

In many cases,it not only need class labels from a clustering algorithm,but also need to understand the cluster structure and data distribution of a dataset.To meet the later demand,many visualization methods are developed to display a multi-dimensional dataset and its clustering result in a two or three dimensional space.This survey introduces more than thirty popular visualization methods of clustering results from three points of view: data and clustering results are displayed directly in a two or three dimensional space,clustering results are displayed after data dimensions are reduced(or data are projected),and other display modes of clustering results.The advantages,disadvantages and applicability of these methods are discussed.