finding and replacing 'nan' with a number

Question

I want to replace number 3 instead of all 'nan' in array. this is my code:

train= train.replace("nan",int(3))

But nothing changes in my array. Could u please guide me?

Sen

Aside: what does print(type(train)) return? Neither lists nor numpy ndarrays have a replace method, so I'm not sure why you're not getting an AttributeError. Is train a pandas Series? – Lipchitz

To follow up on @DSM's comment, if train were a pandas dataframe or Series, there's a fillna() method that handles this sort of situation. – Genevieve

Answer 1

>>> import math
>>> train = [10, float('NaN'), 20, float('NaN'), 30]
>>> train = [3 if math.isnan(x) else x for x in train]
>>> train
[10, 3, 20, 3, 30]

Pearlene

Answer 2

You can use np.isnan:

import numpy as np
train = np.array([2, 4, 4, 8, 32, np.NaN, 12, np.NaN]) 
train[np.isnan(train)]=3
train

Output:

array([  2.,   4.,   4.,   8.,  32.,   3.,  12.,   3.])

Constanceconstancia

Answer 3

This code changes all nan to 3:

y = np.nan_to_num(x) + np.isnan(x)*3

This code changes all 3 to nan:

y = x*(x!=3) + 0/(x!=3)

These methods only work with numpy arrays. Please note that np.nan_to_num(x) always changes nan to 0.

Vadim

Answer 4

The following works with floats, but could not quickly make it with integers... Terveisin Markus

import numpy as np
a=np.array([1,2,3,np.nan])
b=np.nan_to_num(a,nan=3)
print(f"a: {a}")
print(f"b: {b}")

a: [ 1. 2. 3. nan]

b: [1. 2. 3. 3.]

Deify