来自http://bbs.pinggu.org/thread-2448170-1-1.html
原始数据：
CUSTOMER TIME PRODUCT
0 0 hering
0 1 corned_b
0 2 olives
0 3 ham
0 4 turkey
0 5 bourbon
0 6 ice_crea
1 0 baguette
1 1 soda
1 2 hering
1 3 cracker
数据结构如上，共有三个字段：CUSTOMER TIME PRODUCT
libname emlib ‘C:\Users\Administrator\Desktop\emlib’;
proc dmdb batch data=emlib.assocs out=dmassoc dmdbcat=catassoc;
id customer time;
class product(desc);
run;
proc assoc data=emlib.assocs dmdbcat=catassoc
out=datassoc(label=’Output from Proc Assoc’)
items=5 support=20;
cust customer;
target product;
run;
proc rulegen in=datassoc
out=datrule(label=’Output from Proc Rulegen’)
minconf=75;
run;
proc print data=datrule;
run;
proc dmdb batch data=emlib.assocs out=dmassoc dmdbcat=catassoc;
id customer time;
class product(desc);
run;
proc assoc data=emlib.assocs dmdbcat=catassoc
out=datassoc(label=’Output from Proc Assoc’)
items=5 support=20;
cust customer;
target product;
run;
proc rulegen in=datassoc
out=datrule(label=’Output from Proc Rulegen’)
minconf=75;
run;
proc print data=datrule;
run;
关联分析代码如上，非常简单，主要用到两个过程PROC ASSOC和PROC RULEGEN1. PROC ASSOC 主要的作用为生成所有的K-项集，并统计其频率。格式如下：
PROC ASSOC <option(s)>;
CUSTOMER variable-list;
TARGET variable;
CUSTOMER variable-list;
TARGET variable;
option中最重要的两个分别为items=5 和support=20分别代表K项集的最大项目数阈值和支持度（support）
CUST statement 与TARGET statement分别代表你的标志变量和目标变量。
使用该过程的时候需要注意一点，引用SAS官方原文：
CUST statement 与TARGET statement分别代表你的标志变量和目标变量。
使用该过程的时候需要注意一点，引用SAS官方原文：
Processing an extremely large number of sets could cause your system to run out of disk and/or memory resources. However, by using a higher support level, you can reduce the item sets to a more manageable number.
如果你处理的是大数据，运行这个程序的时候可能会磁盘满或者内存溢出。为了避免这种情况的发生，在第一次运行的时候应该设置一个比较高的置信度。逐次递减。
2.PROC RULEGEN 主要的作用是生成关联规则，格式也非常的简单，如下：
2.PROC RULEGEN 主要的作用是生成关联规则，格式也非常的简单，如下：
PROC RULEGEN <option(s)>;
option主要是设置minconf：最小置信度（confidence）
3.运行后生成结果（部分）为：主要有：support（支持度）、confidence（置信度）、lift（梯度），rule（规则）
序列关联：
proc dmdb batch data=emlib.assocs out=dmassoc dmdbcat=catassoc;
id customer time;
class product(desc);
run;
proc sort data=datrule;
by descending lift;
run;
proc print data=datrule(obs=5) label;
var set_size exp_conf conf support lift count
rule _lhand _rhand;
title ‘Top Ten Rules based on Lift’;run;
proc sequence data=dmassoc dmdbcat=catassoc
assoc=datassoc out=sout(label=’output from proc sequence’)
nitems=4;
cust customer;
target product;
visit time/same=2;
run;
id customer time;
class product(desc);
run;
proc sort data=datrule;
by descending lift;
run;
proc print data=datrule(obs=5) label;
var set_size exp_conf conf support lift count
rule _lhand _rhand;
title ‘Top Ten Rules based on Lift’;run;
proc sequence data=dmassoc dmdbcat=catassoc
assoc=datassoc out=sout(label=’output from proc sequence’)
nitems=4;
cust customer;
target product;
visit time/same=2;
run;
时序关联唯一的区别是PROC DMDB（ps：这个过程在SAS/EM非常重要，基本所有数据挖掘过程运行前都必须运行这个过程。主要作用是为建模创建数据仓库）过程中加入标识变量为custom time。时间也加入标识变量之一。关键过程为PROC SEQUENCE
格式如下：
格式如下：
PROC SEQUENCE <option(s)>;
CUSTOMER variable(s);
TARGET variable;
VISIT variable /<visit-option(s)>;
CUSTOMER variable(s);
TARGET variable;
VISIT variable /<visit-option(s)>;
一起大数据提示您：
visit-option SAME and WINDOW specify the upper and lower timing limits of a sequence rule.SAME < time difference WINDOW. Visit-option can be as follows: SAME=same-number Specifies the lower time-limit between the occurrence of two events that you want to associate with each other. If the time difference between the two events is less than or equal to same-number (that is, it is ‘too soon’), then the two events are treated as occurring in the same visit, and the transaction is not counted. Default: 0 WINDOW=window-number Specifies the maximum time difference between the occurrence of two events that you want to be treated as the same visit. If the time difference is greater than window-number (that is, it is ‘too late’), then the transaction is treated as falling outside of the timing window, and the transaction is not counted. For NITEM-long sequence chain, WINDOW applies to the entire chain. Default: MAX
VISIT有两个参数，SAME 定义最小时间间隔，默认为0，也就是没次访问都看成独立的一次，如果设置为1，那么访问序列 1，2，3，4，5会被整合成 （1,2）（3,4）（5）三次。
WINDOW 定义最大的时间间隔，如果设置为3，那么访问序列1,5，不会生成任何有价值的序列，因为时间间隔太久了。
详情请参考 http://support.sas.com/documentation/onlinedoc/miner/em43/sequence.pdf
原创文章，作者：xsmile，如若转载，请注明出处：http://www.17bigdata.com/%e5%a6%82%e4%bd%95%e7%bc%96%e7%a8%8b%e8%b0%83%e7%94%a8em4-3%e4%b8%ad%e7%9a%84%e5%85%b3%e8%81%94%e5%88%86%e6%9e%90%e4%b8%8e%e6%97%b6%e5%ba%8f%e5%85%b3%e8%81%94%e8%bf%87%e7%a8%8b/
更多内容请访问：IT源点
注意：本文归作者所有，未经作者允许，不得转载
