Pandas groupby 多列独占

Question

I have the DataFrame below and want to find the count of y and n for each column:

ID	var1	var2
1	y
2	n	y
3	y	n
4	y	n
5		y

the result would be like this:

	var1_N	var2_N
y	3	2
n	1	2

I used transform function but was wondering there is a better way to get the results. Thanks!

Answer 1

You can just do value_counts on all columns you need to count using apply method, the results will be automatically joined on the index (or var value in your case):

df.filter(like='var').apply(lambda s: s.value_counts())

   var1  var2
y     3     2
n     1     2

Or use pd.value_counts directly:

df.filter(like='var').apply(pd.value_counts)

   var1  var2
y     3     2
n     1     2

Answer 2

You can use melt to flatten your dataframe then use value_counts and unstack the variable column:

>>> df.melt('ID').value_counts(['variable', 'value']).unstack('variable')

variable  var1  var2
value               
n            1     2
y            3     2

You can remove index and column names: by appending ``

Output:

>>> df.melt('ID').value_counts(['variable', 'value']).unstack('variable') \
      .rename_axis(index=None, columns=None)

   var1  var2
n     1     2
y     3     2