CODEWITHSUNDEEP
pythonpandas
MBguitarburst
MBguitarburst

Reputation: 277

Pandas Group By then Merge Using Multiple Columns

I'm pretty new to Pandas but I'm trying to analyze a dataset of employee timestamps to determine the sum of unique daily timestamps per week.

My initial dataframe (input1) looks like this (but much longer):

            ID          Datetime        Week/Year
0          15.0    2019-02-04 08:28:44   6/2019
1          15.0    2019-02-04 12:48:05   6/2019
2          15.0    2019-02-04 12:54:29   6/2019
3          15.0    2019-02-05 08:05:51   6/2019
4          15.0    2019-02-05 12:47:26   6/2019
5          15.0    2019-02-05 14:45:34   6/2019
6          15.0    2019-02-06 08:10:59   6/2019
7          15.0    2019-02-06 12:49:24   6/2019
8          15.0    2019-02-06 13:02:48   6/2019
9          15.0    2019-02-07 08:02:22   6/2019
10         15.0    2019-02-08 08:02:10   6/2019
11         15.0    2019-02-08 09:55:22   6/2019

I created another dataframe:

df = pd.DataFrame({'Timestamp':  input1['Datetime'], 'ID': input1['ID'], 'Week/Year': input1['Week/Year'],'MDY':input1['Server Date/Time'].apply(lambda x: "%d/%d/%d" % (x.month, x.day, x.year))})

Then I grouped by Week, Employee, and got unique count per day (MDY):

df_grouped = df.groupby(['Week/Year', 'ID']).MDY.nunique()

Week/Year   ID    MDY 
6/2019      15.0   5

The end result I'm looking for is to merge the MDY sums back to the initial dataframe by joining on Week and ID - I tried a few different ways:

input1.merge(df_grouped.to_frame(), left_on=['ID','Week/Year'], right_index=True)

to get something like:

           ID          Datetime        Week/Year    MDY
0          15.0    2019-02-04 08:28:44   6/2019    5
1          15.0    2019-02-04 12:48:05   6/2019    5
2          15.0    2019-02-04 12:54:29   6/2019    5
3          15.0    2019-02-05 08:05:51   6/2019    5
4          15.0    2019-02-05 12:47:26   6/2019    5
5          15.0    2019-02-05 14:45:34   6/2019    5

After the join I just end up getting NaN across the board. Anyone able to steer me in the right direction?

Thanks.

Upvotes: 0

Views: 1419

Answers (1)

Andy L.
Andy L.

Reputation: 25269

this groupby

df_grouped = df.groupby(['Week/Year', 'WD: Employee ID']).MDY.nunique()

should return a series has index as Week/Year WD: Employee ID

Week/Year   WD: Employee ID
6/2019      15.0   5
Name: MDY , dtype: int64

However, you show its index as Week/Year ID. You may check columns name to make sure it match.

Next, on this

input1.merge(df_grouped.to_frame(), left_on=['ID','Week/Year'], right_index=True)

Assume df_grouped has index as you show in example which is Week/Year ID, you have left_on wrong order against right_index. It should be

input1.merge(df_grouped.to_frame(), left_on=['Week/Year', 'ID'], right_index=True)

Upvotes: 1

Related Questions