Pandas groupby value_count filter by frequency

问题

I would like to filter out the frequencies that are less than n, in my case n is 2

df = pd.DataFrame({'A' : ['foo', 'bar', 'foo', 'bar', 'foo', 'bar','foo', 'bar', 'foo', 'bar',],'B' : ['yes', 'no', 'yes', 'no', 'no', 'yes','yes', 'no', 'no', 'no']})
df.groupby('A')['B'].value_counts()

A    B  
bar  no     4
     yes    1
foo  yes    3
     no     2
Name: B, dtype: int64

Ideally I would like the results in a dataframe showing the below(frequency of 1 is not excluded)

A    B      freq
bar  no     4
foo  yes    3
foo  no     2

I have tried

df.groupby('A')['B'].filter(lambda x: len(x) > 1)

but this fails as apparently groupby returns a serie

回答1:

This can be down with one line with .loc

df.groupby('A')['B'].value_counts().loc[lambda x : x>1].reset_index(name='count')
Out[530]: 
     A    B  count
0  bar   no      4
1  foo  yes      3
2  foo   no      2

回答2:

You can just store the value_counts output and then just filter it:

In[3]:
counts = df.groupby('A')['B'].value_counts()
counts[counts>=2]

Out[3]: 
A    B  
bar  no     4
foo  yes    3
     no     2
Name: B, dtype: int64

If you want to get your desired output, you can call reset_index and rename the new column:

In[21]:
counts[counts>=2].reset_index(name='count')

Out[21]: 
     A    B  count
0  bar   no      4
1  foo  yes      3
2  foo   no      2

来源：https://stackoverflow.com/questions/50117068/pandas-groupby-value-count-filter-by-frequency