SQL> select * from t1_ order by a,b;

A B
- -
a w
a w
a x
a y
b x
b z
b z
b z
c x
c x
c y

A B
- -
d y
d z

13 rows selected.

SQL> 
SQL> -- Get the count group by a,b first
SQL> -- and then sum them up, order them and pick the first row in each partition to return.
SQL> with v1 as (select a, b, count(1) cnt from t1_ group by a,b order by 1,3 desc)
  2      select a, b, sum(cnt) over (partition by a) total, cnt, 
  3                   row_number() over (partition by a order by cnt desc) rnk from v1;

A B      TOTAL        CNT        RNK
- - ---------- ---------- ----------
a w          4          2          1
a x          4          1          2
a y          4          1          3
b z          4          3          1
b x          4          1          2
c x          3          2          1
c y          3          1          2
d y          2          1          1
d z          2          1          2

9 rows selected.

SQL> 
SQL> with v1 as (select a, b, count(1) cnt from t1_ group by a,b order by 1,3 desc)
  2  select a, total, b, cnt from (
  3      select a, b, sum(cnt) over (partition by a) total, cnt,
  4                   row_number() over (partition by a order by cnt desc) rnk from v1)
  5  where rnk=1;

A      TOTAL B        CNT                                                       
- ---------- - ----------                                                       
a          4 w          2                                                       
b          4 z          3                                                       
c          3 x          2                                                       
d          2 y          1                                                       

SQL> spool off