标题: 用SAS做聚类分析 [打印本页] 作者: shiyiming 时间: 2010-6-28 14:42 标题: 用SAS做聚类分析 请教下各位高手
有没有人用SAS做过聚类分析？如果变量是二分类的变量是不是要选择特定的距离，是什么OPTION呢？
谢谢！作者: shiyiming 时间: 2010-7-1 17:45 标题: Re: 用SAS做聚类分析 It seems that SAS dosen't have any good function to deal with binary variables( or nominal or ordinal variables). Statistical models in SAS CLUSTER Proc can only deal with continual variables. If you want a reasonable result, you'd better transform your nomial data to continual data.作者: shiyiming 时间: 2010-7-2 19:57 标题: Re: 用SAS做聚类分析 赞成gagoon的观点作者: shiyiming 时间: 2010-7-3 07:17 标题: Re: 用SAS做聚类分析 to gzgoon
1. You are wrong. Check Detail section of PROC CLUSTER manual
2. PROC DISTANCE is readily helpful, too作者: shiyiming 时间: 2010-7-5 16:19 标题: Re: 用SAS做聚类分析 to oloolo:
1.Proc distance indeed can deal with categrical or ordianl data, but actually it just transforms them into contunial variables before calculating distance using some very routine methods. If we want a good cluster result, we need do transformation by ourselves.
2.Now there are some papers about how to do cluster when data has categorical or ordinal variables. they are called cluster aggregation. But I haven't find something like this in SAS.作者: shiyiming 时间: 2010-7-6 05:29 标题: Re: 用SAS做聚类分析 to gzgoon
if you r looking for fresh algorithms, you better end up with R, or, implementing your own in SAS