聚类分析的结果分析和编程

《聚类分析的结果分析和编程》由会员上传分享，免费在线阅读，更多相关内容在工程资料-天天文库。

1、聚类分析方法一：最短距离法程序：procclusterdata=work•zidanouttree=treemethod=sinnotie;idg；proctreehorizontal;idg;run;结果:EigenvalueDifferenceProportionCumulative129740435.825650007.90.8718().871824090427.93812173.1011990.99173278254.8274011.40.00820.999944243.44242.50.00011.000050

2、.90.90.00001.000060.00.0().()()()()1.000()70.00.00001.0000ClusterHistoryNormMinNCL-ClustersJoinedFREQDist52620.289541CL530.33333CL4440.500123520.70171CL3CL261.0154方法二：最长距离法程序：procclusterdata=work•zidanouttree=treemethod=complete;idg;proctreehorizontal;idg;run;结果:E

3、igenvaluesoftheCovarianceMatrixEigenvalueDifferenceProportionCumulative129740435.825650007.90.87180.871824090427.93812173.10.11990.99173278254.8274011.40.00820.999944243.44242.50.00011.000050.90.90.00001.000060.00.00.00001.000070.00.00001.0000Root-Mcan-SquarcTotal

4、-SampleStandardDeviation=2207.564MeanDistanceBetweenObservations=7428.48（变量之间的平均距离）ClusterHistoryNormTMaxiNCL■…-ClustersJoinedFREQDiste52620.289541420.500133520.70172CL4CL540.72971CL2CL361.7694此结果表示的是：类的个数，合并类的名字，新类屮的变量个数和最长正规化距离。2'ft0Q2Q4Q60.81.01.2L41.6L8Nfaximu

5、nDistanceBetwenClusters由图知，将1和4分一类，将2和6分一类，将3和5分一类方法三：类平均法程序：procclusterdata=work.zidanouttree=treemethod=average;idg;proctreehorizontal;idg;run;结果：EigenvaluesoftheCovarianceMatrixEigenvalueDifferenceProportionCumulative129740435.825650007.90.87180.871824090427.93

6、812173.1().11990.99173278254.8274011.40.00820.999944243.44242.50.00011.000050.90.90.00001.000060.00.00.00001.000070.0().()()()()1.0(X)0Root-Mean-SquareTotal-SampleStandardDeviation=2207.564Root-Mcan-SquarcDistanceBbtweenCbscrvations=8259.947ClusterHistoryNCL-Clust

7、ersJ

8、eehorizontal;idg;run;结果:EigenvaluesoftheCovarianceMatrixEigenvalueDifferenceProportionCumulative129740435.825650007.90.87180.871824090427.93812173.10」19