Get most common column for each column value

Question

I want the most common letter for each number. I've tried a variety of things; not sure what's the right way.

import pandas as pd
from pandas import DataFrame, Series

original = DataFrame({
    'letter': {0: 'A', 1: 'A', 2: 'A', 3: 'B', 4: 'B'}, 
    'number': {0: '01', 1: '01', 2: '02', 3: '02', 4: '02'}
})

expected = DataFrame({'most_common_letter': {'01': 'A', '02': 'B'}})

Ideally I'm looking to maximize readability.

jezrael · Answer 1 · 2017-06-07T05:33:45.457

Use groupby + apply + value_counts + select first index values, because values are sorted.

Last convert Series to_frame and remove index name by rename_axis:

df = original.groupby('number')['letter'] \
             .apply(lambda x: x.value_counts().index[0])
             .to_frame('most_common_letter')
             .rename_axis(None)
print (df)
   most_common_letter
01                  A
02                  B

Similar solution:

from collections import Counter

df = original.groupby('number')['letter'] \
             .apply(lambda x: Counter(x).most_common(1)[0][0]) \
             .to_frame('most_common_letter') \
             .rename_axis(None)
print (df)
   most_common_letter
01                  A
02                  B

Or use Series.mode:

df = original.groupby('number')['letter'] \
             .apply(lambda x: x.mode()[0][0])
             .to_frame('most_common_letter')
             .rename_axis(None)
print (df)
   most_common_letter
01                  A
02                  B

Is there a way to increase readability by using `.idxmax()` or something? — Hatshepsut, Jun 07 '17 at 05:23

score 2 · Accepted Answer · answered Jun 07 '17 at 07:41

2

We can use DataFrame.mode() method:

In [43]: df.groupby('number')[['letter']] \
           .apply(lambda x: x.mode()) \
           .reset_index(level=1, drop=True)
Out[43]:
       letter
number
01          A
02          B

answered Jun 07 '17 at 07:41

MaxU - stand with Ukraine

205,989
36
386
419

yobogoya · Answer 3 · 2017-06-07T05:38:50.603

>>> df = pd.DataFrame({
    'letter': {0: 'A', 1: 'A', 2: 'A', 3: 'B', 4: 'B'}, 
    'number': {0: '01', 1: '01', 2: '02', 3: '02', 4: '02'}})
>>> df['most_common_letter']=df.groupby('number')['letter'].transform(max) 
>>> df = df.iloc[:,1:].drop_duplicates().set_index('number')
>>> df.index.name = None
>>> df
   most_common_letter
01                  A
02                  B

Or this way if it helps readability:

>>> df['most_common_letter']=df.groupby('number')['letter'].transform(max) 
>>> df = df.drop('letter', axis=1).drop_duplicates().rename({'number': None}).set_index('number')
>>> df
   most_common_letter
01                  A
02                  B

Get most common column for each column value

3 Answers3