Pandas group by a specific value in any of given columns

user_14259461 0
Viewed 2352

Given the pandas dataframe as follows:

   Partner1 Partner2    Interactions
0  Ann      Alice       1
1  Alice    Kate        8
2  Kate     Tony        9
3  Tony     Ann         2

How can I group by a specific partner, let's say to find the total number of interactions of Ann?

Something like

gb = df.groupby(['Partner1'] or ['Partner2']).agg({'Interactions': 'sum'})

and getting the answer:

Partner Interactions
Ann     3
Alice   9
Kate    17
Tony    11
pandas-groupby pandas dataframe python
2 Answers

You can use melt together with groupby. First melt:

df = pd.melt(df, id_vars='Interactions', value_vars=['Partner1', 'Partner2'], value_name='Partner')

This will give:

   Interactions   variable  Partner
0             1   Partner1      Ann
1             8   Partner1    Alice
2             9   Partner1     Kate
3             2   Partner1     Tony
4             1   Partner2    Alice
5             8   Partner2     Kate
6             9   Partner2     Tony
7             2   Partner2      Ann

Now, group by Partner and sum:

df.groupby('Partner')[['Interactions']].sum()

Result:

Partner  Interactions
  Alice             9
    Ann             3
   Kate            17
   Tony            11
user_75795470

You can do merge dataframe itself:

# join the df to itself
join_df = df.merge(df, left_on='Partner1', right_on='Partner2', suffixes=('', '_'))

# get sum
join_df['InteractionsSum'] = join_df[['Interactions', 'Interactions_']].agg(sum, 1)

join_df = join_df[['Partner1', 'Interactions']].copy()

print(join_df)

  Partner1  Interactions
0      Ann             1
1    Alice             8
2     Kate             9
3     Tony             2
user_92992590
Related