성능을 고려하지 않고 만들었습니다.

create table tb_class (
 seq number,
 title varchar2(10)
);

create table tb_class_user (
  seq number,
  user_name varchar2(10),
  user_ci varchar2(10),
  class_seq number
);

delete from tb_class;

insert into tb_class values (1, 'korean');
insert into tb_class values (2, 'english');
insert into tb_class values (3, 'mathematic');

select * from tb_class;

delete from tb_class_user;

insert into tb_class_user values (1, 'aa', '1111', 1);
insert into tb_class_user values (2, 'bb', null, 1);
insert into tb_class_user values (3, 'cc', '3333', 1);

insert into tb_class_user values (4, null, '1111', 2);
insert into tb_class_user values (5, 'bb', '2222', 2);
insert into tb_class_user values (6, 'cc', '3333', 2);

insert into tb_class_user values (7, 'dd', '1111', 3);
insert into tb_class_user values (8, 'cc',  null, 3);


select * from tb_class_user;

    select a.user_name, a.user_ci, b.title, count(1) over(partition by a.user_ci) cnt
     from tb_class_user a
         , tb_class b
  where a.class_seq = b.seq
  and a.user_ci is not null
  ;

-- 원하는 결과
select nvl(max(user_name), '') user_name,
      listagg(title, ',') within group(order by title) title,
      user_ci
 from (
    select a.user_name, a.user_ci, b.title, count(1) over(partition by a.user_ci) cnt
     from tb_class_user a
         , tb_class b
  where a.class_seq = b.seq
    and a.user_ci is not null
  )
  where cnt > 2
 group by user_ci;
 

user_ci 가 null 인 자료가 들어오는게 가능한 건가요?
- 업무 로직 검토 필요합니다.
- 가능하다면 왜 그런지 상세 설명이 필요합니다.
Group By 에 a.name 을 추가하면 MAX 필요 없습니다.
NVL 은 굳이 필요 없을 듯 하구요.
Distinct 가 필요한지도 검토 필요합니다.
wm_concat 는 비공식 함수입니다. 11G 라면 ListAgg 이용하세요.
 

SELECT a.user_ci
     , a.name
     , WM_CONCAT(b.title) class_list_10g
     , LISTAGG(b.title, ',') WITHIN GROUP(ORDER BY b.seq) class_list_11g
  FROM tb_class_user a
     , tb_class b
 WHERE a.class_seq = b.seq
 GROUP BY a.user_ci, a.name
--HAVING COUNT(DISTINCT a.class_seq) > 2
--HAVING COUNT(a.class_seq) > 2
HAVING COUNT(*) > 2
;