Need to Calculate few metrics from dataset using SQL - separate queries - mysql

Question

Dataset looks like this : This is a sample dataset for number of employee login activity named - activity
I need to calculate few metrics, was able to do in python data frames, but new in mySQL.
what is the average number of employee active per day for month of jan 2018 by dept ( was able to do somewhat half of it, but results coming are not correct.
number of unique active employee (login >0) per month for jan 2018 for each dept_id (was able to do it)
month over month growth for all dept_id from dec-2017 to jan 2018 where at least one employee was active (login >0) - no idea how to do this in sql
fraction of users who were active in each dept_id for dec 2017 and were also active in the same dept_id for jan 2018
how many employee login in on 3 or more consecutive days in jan 2018
Any help would be appreciated.
Query written for case 1:
select dept_id,
DAU
from
(
select dept_id
, month(date)
, year(date)
, avg(logins) as DAU
from
(select * from activity where login >0)
where year(date) =2018
and month(date) =1
group by dept_id, month(date), year(date)
)
Textual Format Dataset
date dept_id emp_id logins
29-11-2017 ABC001 A1 1
30-11-2017 ABC002 A2 2
01-12-2017 XYZ001 A3 0
01-12-2017 XYZ002 A4 1
03-12-2017 ABC001 D2 4
04-12-2017 ABC002 D1 1
05-12-2017 XYZ001 A6 2
05-12-2017 XYZ002 A7 3
30-12-2017 ABC001 A8 0
01-01-2018 ABC002 A2 6
02-01-2018 XYZ001 A10 4
03-01-2018 XYZ002 A11 2
04-01-2018 ABC001 A1 2
04-01-2018 ABC002 A2 0
05-01-2018 XYZ001 A13 4
05-01-2018 XYZ001 A6 2
05-01-2018 XYZ002 A7 1
06-01-2018 XYZ001 A6 2
06-01-2018 XYZ002 A7 3
07-01-2018 XYZ001 A6 3
07-01-2018 XYZ002 A7 4
06-01-2018 XYZ002 A14 3
30-01-2018 ABC001 A15 2

Answer 1

Let me know if this works otherwise I will update the answer, I don't have MYSQL installed so wasn't able to check.
And the date is a keyword in oracle but not sure in MYSQL so use it in quotes like "date".
Case 1:
SELECT dept_id,
AVG(cnt) average_emp
FROM (SELECT dept_id,
days,
COUNT(emp_id) cnt
FROM (SELECT dept_id,
emp_id,
SUM(logins) logins,
to_char(DATES, 'dd') days
FROM mytable
WHERE to_char(DATES,'mmyyyy') = '012018'
GROUP BY dept_id,
emp_id,
to_char(DATES, 'dd') )
WHERE logins > 0
GROUP BY dept_id,
days )
GROUP BY dept_id;
Case 2:
SELECT dept_id,
COUNT(emp_id)
FROM (SELECT dept_id,
emp_id,
SUM(logins) logins
FROM mytable
WHERE to_char(DATES,'mmyyyy') = '012018'
GROUP BY dept_id,
emp_id )
WHERE logins > 0
GROUP BY dept_id;
Case 3:
SELECT months,
users,
ROUND( (users - nvl(LAG(users) OVER (ORDER BY rownum),users) ) / nvl(LAG(users) OVER (ORDER BY ROWNUM), 1)
, 2) growth_rate
FROM (SELECT to_char(mt.DATES, 'MON-YYYY') months,
count(mt.EMP_ID) users
FROM (SELECT *
FROM MYTABLE
ORDER BY DATES) mt
WHERE mt.DATES >= to_date('DEC-2017', 'MON-YYYY')
AND mt.DATES <= to_date('JAN-2018', 'MON-YYYY')
GROUP BY to_char(mt.DATES, 'MON-YYYY')
ORDER BY to_date(months, 'MON-YYYY') ) oq
WHERE exists(SELECT 1
FROM MYTABLE iq
WHERE to_char(iq.DATES, 'MON-YYYY') = oq.months
AND iq.EMP_ID IN (SELECT EMP_ID
FROM MYTABLE
WHERE iq.LOGINS > 0) );
Case 4:
SELECT dept_id,
emp_id
FROM (SELECT dept_id,
emp_id
FROM mytable
WHERE to_char(DATES,'mmyyyy') = '122017'
AND logins > 0
GROUP BY dept_id,
emp_id )
INTERSECT
SELECT dept_id,
emp_id
FROM (SELECT dept_id,
emp_id
FROM mytable
WHERE to_char(DATES,'mmyyyy') = '012018'
AND logins > 0
GROUP BY dept_id,
emp_id )
Case 5:
-- not full proof
SELECT COUNT(*) emp_cnt
FROM (SELECT emp_id,
DENSE_RANK() OVER(ORDER BY DATES) rn,
COUNT(*) OVER(PARTITION BY emp_id ORDER BY DATES) cnt
FROM mytable
WHERE to_char(DATES,'mmyyyy') = '012018'
AND logins > 0
ORDER BY rn,
cnt )
WHERE rn = cnt
AND rn >= 3;