Pandas - 通过聚合非空值替换NaN

Question

Suppose I have a DataFrame with some NaN -

import pandas as pd
l = [{'C1':-6,'C3':2},
     {'C2':-6,'C3':3},
     {'C1':-6.3,'C2':8,'C3':9},
     {'C2':-7}]
df1 = pd.DataFrame(l,
    index=['R1','R2','R3','R4'])
print(df1)

     C1   C2   C3
R1 -6.0  NaN  2.0
R2  NaN -6.0  3.0
R3 -6.3  8.0  9.0
R4  NaN -7.0  NaN

Problem - If there is any NaN value in any row cell then it has to be replaced by the aggregate of non-null values from the same row. For instance, in first row, the value of (R1,C2) should be = (-6+2)/2 = -2

Expected output -

     C1   C2   C3
R1 -6.0 -4.0  2.0
R2 -1.5 -6.0  3.0
R3 -6.3  8.0  9.0
R4 -7.0 -7.0 -7.0

Answer 1

Use apply with axis=1 for process by rows:

df1 = df1.apply(lambda x: x.fillna(x.mean()), axis=1)
print(df1)

     C1   C2   C3
R1 -6.0 -2.0  2.0
R2 -1.5 -6.0  3.0
R3 -6.3  8.0  9.0
R4 -7.0 -7.0 -7.0

Also works:

df1 = df1.T.fillna(df1.mean(1)).T
print(df1)
     C1   C2   C3
R1 -6.0 -2.0  2.0
R2 -1.5 -6.0  3.0
R3 -6.3  8.0  9.0
R4 -7.0 -7.0 -7.0

Because:

df1 = df1.fillna(df1.mean(1), axis=1)
print(df1)

NotImplementedError: Currently only can fill with dict/Series column by column

Answer 2

You could do this. Transpose, then do fillna() then again transpose it.

>>> df1 = df1.T.fillna(df1.mean(axis=1)).T
>>> print(df1)
     C1   C2   C3
R1 -6.0 -4.0  2.0
R2 -1.5 -6.0  3.0
R3 -6.3  8.0  9.0
R4 -7.0 -7.0 -7.0