再用K-means 算法聚类之后判断出可以聚成3类，接下来的问题是要将原始数据降到两维，通过图形探测整个数据聚类后大致大类别数。但我的问题是以下这两种方法用哪个比较好？求高人指点。谢谢！

cannonical discriminant analysis
proc candisc data=outdata3 out=clustcan ncan=2;
class cluster;
var alcevr1 marever1 alcprobs1 deviant1 viol1 dep1 esteem1 schconn1
    parpres paractv famconct;
run;


principle component analysis
proc princomp data=outdata3 out=clustprin n=2;
var alcevr1 marever1 alcprobs1 deviant1 viol1 dep1 esteem1 schconn1
    parpres paractv famconct;
run;


proc sgplot data=clustcan;
scatter y=can2 x=can1/group=cluster;
run;
quit;


proc sgplot data=clustprin;
scatter y=prin2 x=prin1/group=cluster;
run;
quit;