pandas group by and add columns data based on conditions

Question

I have merged test data as below:

Device       time    Key score
Computers 2018-01-01 14.0 4.0
Computers 2018-01-01 11.0 4.0
Computers 2018-01-01 16.0 0.0

I need to group data by columns [Device,time] and by max value from column score and get minimum key value assigned to this score.

My 1 atempt:

df_out = df_out.groupby(['Device', 'time'])['score'].max().reset_index()

Output 1:

Device       time    score
Computers 2018-01-01 4.0

My 2 atempt:

df_out = df_out.groupby(['Device', 'time'])['score', 'Key'].max().reset_index()

Output 2:

Device       time    score Key
Computers 2018-01-01 4.0  14.0

How to get assigned proper minimum Key?

Desired output:

Device       time    score Key
Computers 2018-01-01 4.0  11.0

Thanks for You hard work.

Answer 1

You can use transform :

df[df.score.eq(df.groupby(['Device', 'time'])['score'].transform('max'))]

---

      Device        time   Key  score
0  Computers  2018-01-01  14.0    4.0

As per EDIT:

df.groupby(['Device', 'time'],as_index=False).agg({'score':'max','Key':'min'})

---

      Device        time  score   Key
0  Computers  2018-01-01    4.0  11.0

Answer 2

Using apply and custom function to get the desired row with loc :

def selecting(x):
    subx = x.loc[x['score'] == x['score'].max()]
    return subx.loc[subx['Key'].idxmin()]

ddf = df.groupby(['Device', 'time']).apply(selecting)

Using your sample input, this will give:

1                        Device        time   Key  score
Device    time                                          
Computers 2018-01-01  Computers  2018-01-01  11.0    4.0

You can drop the multi-index using .reset_index(drop=True) on the result.

I edited the answer using a custom function, to perform correctly the selection. I realized that the previous version of my answer may raise a KeyError on more complex dataframes.