/* Демонстрационный набор тестовых данных , тут I,i - вход, O,o - выход*/ 
with t as 
  (
    select 1 as user_id, to_date('2017-11-01','yyyy-mm-dd') as x_dt, 'I' as event_type union all
    select 1 as user_id, to_date('2017-11-02','yyyy-mm-dd') as x_dt, 'I' as event_type union all
    select 1 as user_id, to_date('2017-11-03','yyyy-mm-dd') as x_dt, 'O' as event_type union all
    select 1 as user_id, to_date('2017-11-04','yyyy-mm-dd') as x_dt, 'O' as event_type union all
    select 2 as user_id, to_date('2017-11-05','yyyy-mm-dd') as x_dt, 'I' as event_type union all
    select 2 as user_id, to_date('2017-11-06','yyyy-mm-dd') as x_dt, 'O' as event_type union all
    select 2 as user_id, to_date('2017-11-07','yyyy-mm-dd') as x_dt, 'I' as event_type union all
    select 2 as user_id, to_date('2017-11-08','yyyy-mm-dd') as x_dt, 'O' as event_type union all
    select 3 as user_id, to_date('2017-11-09','yyyy-mm-dd') as x_dt, 'O' as event_type union all
    select 3 as user_id, to_date('2017-11-10','yyyy-mm-dd') as x_dt, 'I' as event_type union all
    select 3 as user_id, to_date('2017-11-11','yyyy-mm-dd') as x_dt, 'I' as event_type union all
    select 3 as user_id, to_date('2017-11-12','yyyy-mm-dd') as x_dt, 'O' as event_type union all
    select 4 as user_id, to_date('2017-11-13','yyyy-mm-dd') as x_dt, 'I' as event_type union all
    select 5 as user_id, to_date('2017-11-14','yyyy-mm-dd') as x_dt, 'O' as event_type
  )
/* Возможный вариант решения поставленной задачи*/
select user_id, d1 as date_i, d2 as date_o
from 
  (--  свертка подряд идущих однотипных событий в одно, для входов - поиск ближайшего выхода
    select user_id, event_type, max(x_dt) as d1
         , case when event_type = 'I' and lead(event_type) over(partition by user_id order by min(x_dt)) = 'O'
           then lead(min(x_dt)) over(partition by user_id order by min(x_dt)) end as d2
    from 
      (
        select user_id, x_dt, event_type
             , row_number() over(partition by user_id, event_type order by x_dt) -
               row_number() over(partition by user_id order by x_dt) as x_group
        from   t
      ) dv0
    group by user_id, event_type, x_group
  ) dv1
where event_type = 'I'
order by user_id, date_i