Question

我想实现以下目标：对于表A中的每个客户（custID），获取trade_date并找到小于或等于trade_date的最大创建日期（基于客户作为匹配字段）。获取特定最大创建日期记录的评级，然后重新加入表A（这是表C所需要的）

Trade_Table A：

CustID | Trade_date | Trade_ID
12345  | 30/7/2018  |  4axd
12345  | 30/7/2018  |  
12345  | 31/7/2018  |  5FETF
12345  | 05/9/2018  |  fst43d
12366  | 01/8/2018  |  g3fgg
12377  | 01/9/2018  |  dfd45

风险评分表B：

CustID | Create_date | Rating
12345  | 29/7/2018   |  2
12345  | 30/7/2018   |  3
12345  | 31/7/2018   |  4
12345  | 01/9/2018   |  1
12366  | 30/7/2018   |  1
12377  | 31/9/2018   |  5

最终表C：

CustID | Trade_date | Trade_ID | Rating
12345  | 30/7/2018  |  4axd    | 3
12345  | 30/7/2018  |          | 3
12345  | 31/7/2018  |  5FETF   | 4
12345  | 05/9/2018  |  fst43d  | 1
12366  | 01/8/2018  |  g3fgg   | 1
12377  | 01/9/2018  |  dfd45   |

我尝试了很多方法，这是其中一种，但是我得到了超过1行的返回错误。这比我想的要复杂。

proc sql;
create Final_Table C as select 
A.*, 
(select max(B.Create_date) FROM Risk_Rating_Table B where A.trade_date >=  B.Create_date
group by B.CustID 
)  as Rating
from
Trade_Table A 
;
QUIT;

Answer 1

尝试以下

SELECT Tmp.CustID ,Tmp.Trade_date,Tmp.Trade_ID,Tmp.Create_date,B.Rating 
FROM (    SELECT A.CustID ,A.Trade_date, A.Trade_ID, MAX(B.Create_date) as Create_date
    FROM TableA A
    JOIN TableB B ON A.CustID = B.CustID 
    WHERE A.Trade_date >=B.Create_date
    GROUP BY A.CustID ,A.Trade_date, A.Trade_ID
)Tmp
JOIN TableB B ON Tmp.CustID = B.CustID AND Tmp.Create_date = B.Create_date

Answer 2

正如SQL Guru @Gordon Linoff所说的那样，在PROC SQL中很难做到这一点，我认为最好使用一些数据步骤或哈希技术。我能够效仿您的结果，但不确定其效率如何。我已经转置了数据，并使用数组来避免多对多连接。

 /* create test data*/
  data Trade_Table;
  infile datalines missover;
 input CustID  Trade_date:ddmmyy10.  Trade_ID $;
  format Trade_date:ddmmyy10.;
 datalines;
 12345   30/7/2018    4axd
 12345   30/7/2018    
 12345   31/7/2018    5FETF
 12345   05/9/2018    fst43d
 12366   01/8/2018    g3fgg
 12377   01/9/2018    dfd45
  ;

 /* second test data */     
 data Risk_Rating_Table ;
 input CustID  Create_date:ddmmyy10.  Rating;
 format Create_date ddmmyy10.;
 datalines;
 12345   29/7/2018     2
 12345   30/7/2018     3
 12345   31/7/2018     4
 12366   30/7/2018     1
 12377   30/9/2018     5
  ;

 /* tranpose the data so the date comes in rows and then it is easy to compare*/
  proc transpose data = Risk_Rating_Table out =one(drop =_name_);
   by custid ;
   var create_date;
   run;

   /* merge the master table with tranposed table and may to have sort data*/
   data have1;
   merge Trade_Table one;
   by custid;
    run;

    /* compare ge columns in a array*/
    data have2;
    set have1;
    array col(*) col1-col3;
    do i=1 to 3 while(col{i} le Trade_date);
   new=col{i};
     end;
    format new ddmmyy10.;
     /* try out more examples may for this part*/
     if new = . then new= max(of col:);
     drop col: i;
    run;

    /* get back required columns */
      proc sql;
       create table want as
     select a.CustID ,Trade_date , Trade_ID , case  
        when a.custid = b.custid and trade_date ge create_date
         then b.rating
         else .
         end as rating
    from have2 a
  left join
  Risk_Rating_Table b
   on a.custid = b.custid
   and create_date =new;

Answer 3

我设法按照下面的方法解决此问题-欢迎发表评论。

    proc sql;
create table tmp_dcpbase as SELECT 
A.custid ,A.Trade_date, 
CASE
    WHEN  A.Trade_date >=B.CREATE_DATE is not null  THEN 1
    ELSE  0
END as CONSIDER ,
PUT(max(B.CREATE_DATE), datetime16. ) AS RPQCr8DTForm,
max(B.CREATE_DATE) as RPQCr8DT
FROM Trade_Table as a inner join RISK_RATING_Table as b
ON (a.custID = b.custID ) 
GROUP BY A.custID , A.trade_date,CONSIDER 
having MAX(CONSIDER) = 1
;
quit;

proc sql;
create table target_C as select 
D.custID , D.Trade_date,D.Trade_id, tmp2.RPQCr8DTForm , TMP2.Rating from
(
SELECT A.*, b.Rating
FROM 
tmp_dcpbase A
INNER JOIN RISK_RATING_Table  b ON (a.custID = b.custID AND a.RPQCr8DT = b.CREATE_DATE)
) TMP2 right join trade_table D ON (TMP2.custID =  D.custID AND TMP2.trade_date = d.trade_date)
;
quit;

SAS / SQL-获取日期字段表A大于或等于表B中的日期字段，并从表B中获取最大日期

3 个答案: