Applying non-hierarchical cluster analysis algorithms to climate classification: Some problems and their solution

被引:41
作者
Gerstengarbe, FW
Werner, PC
Fraedrich, K
机构
[1] Potsdam Inst Climate Impact Res, D-14412 Potsdam, Germany
[2] Univ Hamburg, Inst Meteorol, D-20146 Hamburg, Germany
关键词
Cluster Analysis; Cluster Algorithm; Analysis Algorithm; Cluster Number; Practical Relevance;
D O I
10.1007/s007040050118
中图分类号
P4 [大气科学(气象学)];
学科分类号
0706 ; 070601 ;
摘要
Extended non-hierarchical cluster analysis is improved by deriving the initial cluster number and estimating the outliers in the final cluster set. These improvements are tested and compared with an established cluster algorithm using a toy example. Applying the improved cluster analysis to a classification of the European climates shows that the proposed techniques can be of great practical relevance.
引用
收藏
页码:143 / 150
页数:8
相关论文
共 12 条
[1]  
[Anonymous], REV INT STAT I
[2]  
BACHER J, 1996, CLUSTERANALYSE
[3]  
FORGY EW, 1965, BIOMETRICS, V21, P768
[4]   A method to estimate the statistical confidence of cluster separation [J].
Gerstengarbe, FW ;
Werner, PC .
THEORETICAL AND APPLIED CLIMATOLOGY, 1997, 57 (1-2) :103-110
[5]  
JAHNKE H, 1988, CLUSTER ANAL PROCEDU
[6]  
MANN HB, 1947, ANN MATH STAT, V18, P52, DOI DOI 10.1214/AOMS/1177730491
[7]  
Muller P.H., 1973, TAFELN MATH STAT
[8]  
Pettitt A. N., 1979, Applied Statistics, V28, P126, DOI 10.2307/2346729
[9]  
*SAS, 1990, SAS STAT US GUID VER, V1
[10]  
*SPSS, 1999, SPSS WIND MAN