Impute missing values

imputant.py

import numpy as np
from sklearn.preprocessing import Imputer

#generate some data
df1 = np.array(np.random.randn(1000)).reshape(100,10)

#make some values 'NaN'
df1[(df1>-.05) & (df1<.05)] = np.nan

X = df1

print(X)

imp = Imputer(missing_values='NaN', strategy='mean', axis=0)
X = imp.fit_transform(X)

print(X)

the working bits of code are:

imp = Imputer()

Where you create an instance of the Imputer

missing_values='NaN', strategy='mean', axis=0 are all defaults so don't actually need to be stated. It looks at the other values in a column (feature) to generate new values by default (axis = 0)

and

X = imp.fit_transform(X)

Which actually fits the Imputer to the data and creates a new array with the missing values filled