没有下线行为数据,那这次登入是否算登入。没有下线行为数据,如果不算的话，比较好处理。自己写个程序跑下就出来了。

你的统计周期是多长， 24hrs  or 一周 or 更长？ 如果是一天， 讷么用01's 24:00作为cut-off point, 如果你找不到相应的02.

另外的一种方法就是impute. 分类找到这些客户，回填the missing values.

no comments but hope it is clear enough.

proc sort data = rawdata;

by id operate_time type ;

 

data good (drop = lag_type type rename = (operate_time = end_time)) need_work1 need_work2;

format start_time datetime.;

set rawdata ;

by id operate_time type;

lag_type = lag(type);

start_time = lag(operate_time);

if first.id then do;

      lag_type = '';

      start_time = .;

end;

if lag_type = '01' and type = '02' then do;

    format stay time.;

      stay = (operate_time - start_time) ;

    output good;

end;

else if (first.id and type = '02')

     or (type = lag_type) then output need_work1;

else if (last.id and type = '01') then output need_work2;

run;

 

data done_work (keep = id start_time end_time notes);

set need_work1(in = a) need_work2 (in = b);

by id operate_time type;

format notes $ 20. end_time datetime.;

 

if lag_type in ( '', '02') and type = '02' then do;

      notes = 'Missing Start Time';

      end_time = operate_time;

      start_time = .;

end;

 

else if lag_type = '01' and type = '01' then do;

      notes = 'Missing End Time';

end;

 

if b then do;

      notes = 'Non Closed Session';

      start_time = operate_time;

end;

 

run;

 

proc sort data = done_work;

by id start_time;

 

 

data final;

set good (in = g) done_work;

by id start_time;

if g then notes = 'Good Data';

run;