pandas dataframe fillna() not working?

Question

I have a data set in which I am performing a principal components analysis (PCA). I get a ValueError message when I try to transform the data. Below is some of the code:

import pandas as pd
import numpy as np
import matplotlib as mpl
from sklearn.preprocessing import StandardScaler
from sklearn.decomposition import PCA as sklearnPCA

data = pd.read_csv('test.csv',header=0)
X = data.ix[:,0:1000].values   # values of 1000 predictor variables
Y = data.ix[:,1000].values     # values of binary outcome variable
sklearn_pca = sklearnPCA(n_components=2)
X_std = StandardScaler().fit_transform(X)

It is here that I get the following error message:

ValueError: Input contains NaN, infinity or a value too large for dtype('float64').

So I then checked whether the original data set had any NaN values:

print(data.isnull().values.any())   # prints True
data.fillna(0)                      # replace NaN values with 0
print(data.isnull().values.any())   # prints True

I don't understand why data.isnull().values.any() is still printing True even after I replaced the NaN values with 0.

Solution 1

There are two way to achieve, try replace in place:

import pandas as pd

data = pd.DataFrame(data=[0,float('nan'),2,3])   
print('BEFORE:', data.isnull().values.any())   # prints True

# fillna function
data.fillna(0, inplace=True)

print('AFTER:',data.isnull().values.any())   # prints False now :)

Or, use returned object:

data = data.fillna(0)

Both case have same result as following:

BEFORE: True
AFTER: False

Solution 2

You have to replace data by the returned object from fillna

Small reproducer:

import pandas as pd

data = pd.DataFrame(data=[0,float('nan'),2,3])

print(data.isnull().values.any())   # prints True
data = data.fillna(0)                      # replace NaN values with 0
print(data.isnull().values.any())   # prints False now :)