对极端值不敏感的归一化

--创建Rank临时表
CREATE MULTISET VOLATILE TABLE MEMBER_TDTL, NO LOG AS(
    SELECT Member_Id
        ,COUNT(1) AS TC
        ,DENSE_RANK() OVER(ORDER BY TC)AS Tc_Row
    FROM MEMBER_TRADE_PAYDTL
    GROUP BY Member_Id
)WITH DATA  
PRIMARY INDEX(Member_Id,Tc_Row)
ON COMMIT PRESERVE ROWS

--归一化到固定范围分级,以10为例
SELECT Member_Id
    ,TC
    ,CAST(9*(Tc_Row-1)/(1.000*TR.Tc_Row_Max-1)+1 AS INT) AS Tc_Tile
FROM MEMBER_USE_POINT_TDTL,(
    SELECT MAX(Tc_Row) AS Tc_Row_Max
    FROM MEMBER_USE_POINT_TDTL) TR 
  
MEMBER_TRADE_PAYDTL源表就是一些交易订单,一笔交易一个记录
posted @ 2018-11-11 12:01  谢耳朵的派森笔记  阅读(419)  评论(0编辑  收藏  举报