CODEWITHSUNDEEP
pythonpandasdataframe
ah bon
ah bon

Reputation: 10051

Fillna of one column with random float values in range in Pandas

I have such dataframe, how can I fillna with random float values from -0.5 to 0.5 (bounds included) for value column?

  city district       date  value
0    a        b   2019/8/1   0.15
1    a        b   2019/9/1   0.12
2    a        b  2019/10/1    NaN
3    c        d   2019/8/1   0.03
4    c        d   2019/9/1  -0.36
5    c        d  2019/10/1    NaN

Thank you.

Upvotes: 1

Views: 1585

Answers (3)

Aleksandr Iurkin
Aleksandr Iurkin

Reputation: 66

You can use apply method:

def func(val):
    if np.isnan(val):
        return random.uniform(-0.5, 0.5)
    else:
        return val

df['value'] = df['value'].apply(func)

Upvotes: 1

jezrael
jezrael

Reputation: 863226

If want replace each NaNs use Series.mask:

df['value'] = df['value'].mask(df['value'].isna(), np.random.uniform(-0.5, 0.5, size=len(df)))

Or numpy.where:

df['value'] = np.where(df['value'].isna(), 
                       np.random.uniform(-0.5, 0.5, size=len(df)), 
                       df['value'])

print (df)
  city district       date     value
0    a        b   2019/8/1  0.150000
1    a        b   2019/9/1  0.120000
2    a        b  2019/10/1  0.227054
3    c        d   2019/8/1  0.030000
4    c        d   2019/9/1 -0.360000
5    c        d  2019/10/1 -0.001102

Upvotes: 3

ignoring_gravity
ignoring_gravity

Reputation: 10531

You need to specify what you mean by "random". Assuming you mean "uniformly at random", you can use fill_na:

df.value = df.value.fillna(
    pd.Series(np.random.uniform(-0.5, 0.5, size=len(df)), index=df.index)
)

Upvotes: 2

Related Questions