[FIXED] Pandas, Apply function return wrong results

Issue

I've this issue with this function, which must return the proportion of each feature in one column.

Here's some data much easy to make up an idea:

df2 = pd.DataFrame({'X': ['A', 'A', 'B' , 'C'], 'Y': [1, 0, 0 , 1], 'Z': [1, 0, 1 , 1]})

df2['X'].value_counts() 

When I count the values I get

A    2
B    1
C    1

Now, I need to get the proportion for each value of "X"

for freq in df2['X'].value_counts():
    #print(freq)
    print(freq/df2['X'].value_counts().sum())

The result below :

0.5
0.25
0.25

Perfect,

Now I must apply to my Dataframe and get a new column. Below the function:

def get_proportion(df):
    for freq in df2['X'].value_counts():
        return (freq/df2['X'].value_counts().sum())

df2["A"]=df2.apply(get_proportion, axis=1)

result:

    X   Y   Z   A
0   A   1   1   0.5
1   A   0   0   0.5
2   B   0   1   0.5
3   C   1   1   0.5

I should get

    X   Y   Z   A
0   A   1   1   0.5
1   A   0   0   0.5
2   B   0   1   0.25
3   C   1   1   0.25

What's wrong ?

If I set an argument

df2["A"]=df2.apply(get_proportion(df2), axis=1)

I get an error

TypeError: 'numpy.float64' object is not callable

Thank you if you can help.

Solution

df2["A"] = df2.X.apply(lambda x: (df2["X"].value_counts() / len(df2))[x])

---

• len(df2) is number of rows of the dataframe df2,
• (df2["X"].value_counts() / len(df2)) is a series with relative occurences of elements in the column "X".

Answered By - MarianD