如何将数据分组到 5 分钟的用户 bin 中并随后计算记录？

Question

So I have a data frame containing timestamps:

new_date            id  
-------------------  ----  
2021-03-22 00:12:29 164616
2021-03-22 00:11:51 297284
2021-03-22 00:11:19 148817
2021-03-22 00:11:19 139208
2021-03-22 00:10:29 301459
2021-03-22 00:09:48 299543
2021-03-22 00:09:12 302444

I want to split the bins into 5 mins intervals and add together the number of ids of active users that fits withtin the bins.

new_date            id  
-------------------  ----  
2021-03-22 00:20:00 0
2021-03-22 00:15:00 13
2021-03-22 00:10:00 5
2021-03-22 00:05:00 2

so far I have tried

date["new_dates"] = pd.to_datetime(date['\tgp:last_session_date'], errors='coerce')
date = date.drop('\tgp:last_session_date', 1)
date.dropna()
df.groupby(pd.Grouper(key ="new_dates", freq = '5Min')).agg({"\tuser_id": "count"})

But it gives a weird output with different dates.....

2021-02-24 18:45:00 1
2021-02-24 18:50:00 0
2021-02-24 18:55:00 0
2021-02-24 19:00:00 0
2021-02-24 19:05:00 0

Answer 1

I think ouput is expected, if there is some 'lost' datetime near 2021-02-24 18:45:00 .

You can sorting original data for see it:

df = df.sort_values('new_date')

So then this row is count for 1 and next values are 0 , because not exist this datetimes in data (and ouput is consecutive DatetimeIndex)

EDIT:

If need remove NaNs is necessary return back ouput for DataFrame.dropna , else not working (or use alternative):

date["new_dates"] = pd.to_datetime(date['\tgp:last_session_date'], errors='coerce')
date = date.drop('\tgp:last_session_date', 1)
date = date.dropna()
#alternative
#date.dropna(inplace=True)

df = df.sort_values('new_date')
print (df)
df.groupby(pd.Grouper(key ="new_dates", freq = '5Min')).agg({"\tuser_id": "count"})