How to randomly insert NaN in a matrix with NumPy in Python ?

Prerequisites: Numpy

In this article, let’s see how to generate a Python Script that randomly inserts Nan into a matrix using Numpy. Given below are 3 methods to do the same:

Method 1: Using ravel() function

ravel() function returns contiguous flattened array(1D array with all the input-array elements and with the same type as it). A copy is made only if needed.
Syntax :

numpy.ravel(array, order = 'C')

Approach:

• Import module
• Create data
• Choose random indices to Nan value to.
• Pass these indices to ravel() function
• Print data

Example 1:

Python3

import numpy as np
import pandas as pd

# number of nan we want to add It will insert 3 nan values to the data.....
n = 3

# creating dataset
data = np.random.randn(5, 5)

# choosing random indexes to put NaN
index_nan = np.random.choice(data.size, n, replace=False)

# adding nan to the data.
data.ravel()[index_nan] = np.nan
print(data)

Output:

Example 2: Adding nan to but using randint function to create data. For using np.nan in randint function we must first convert the data into float as np.nan is of float type.

Python3

import numpy as np
# number of nan we want to add It will insert 3 nan values to the data.....
n_b = 5

# creating dataset
data_b = np.random.randint(10, 100, size=(5, 5))

# converting the data to float as nan is also of type float
data_b = data_b*0.1

# choosing random indexes to put NaN
index_b = np.random.choice(data_b.size, n_b, replace=False)

# adding nan to the data.
data_b.ravel()[index_b] = np.nan
print(data_b)

Output:

Method 2: Creating mask 

Creating a mask of boolean and applying that mask to the dataset can be one approach to produce the required result.

Approach:

• Import module
• Create data
• Create mask
• Shuffle the mask to randomly apply Nan values
• Apply the mask to the data
• Print data

Example :

Python3

import numpy as np

# creating dataset
X = 10
Y = 5
N = 15

data = np.random.randn(X, Y)

# making a array randomly of same size as data of bool type
mask = np.zeros(X*Y, dtype=bool)

# marking first n indexes as true
mask[:N] = True

# shuffling the mask
np.random.shuffle(mask)
mask = mask.reshape(X, Y)

# applying mask to the data
data[mask] = np.nan
print(data)

Output:

Method 3: Using insert() 

Using insert() function will convert a whole row or a whole column to NaN. This function inserts values along the mentioned axis before the given indices.
Syntax :

numpy.insert(array, object, values, axis = None)

Approach:

• Import module
• Create data
• Use insert Nan values
• Print data

Example:

Python3

import numpy as np

a = np.array([(13.0, 1.0, -47.0), (12.0, 3.0, -47.0), (15.0, 2.0, -44.0)])

# adding nan values to the row
np.insert(a, 2, np.nan, axis=0)

# adding nan values to the row
np.insert(a, 2, np.nan, axis=1)

Output: