How to add a randomly generated number to elements in a pandas DataFrame

Question

I am trying to add/subtract a random number from existing elements (floats) in a pandas DataFrame (Python).

indices is a random subset index, and modify_columns is a list of the columns I wish to modify. My DataFrame is as follows ( active_set.loc[indices,modify_columns] ):

    Values
380977  0.0
683042  0.0
234012  0.0
16517   0.0
... ...

I would like to add or subtract a randomly generated integer (either -1 or 1) from these values.

I have tried using (2*np.random.randint(0,2,size=(count))-1) to generate an array of these random numbers, and add them:

active_set.loc[indices,modify_columns] = active_set.loc[indices,modify_columns] + (2*np.random.randint(0,2,size=(count))-1)

This does not work as there is a ValueError: Unable to coerce to Series, length must be 1: given 180 . I think I can simply create a second DataFrame with the random numbers, or iterate, but these seem inefficient, and there must be a way to use .apply , so I am asking for some help on how to do this.

Answer 1

more general

df.loc[indexes,columns] = df.loc[indexes,columns] + 2*np.random.randint(0,50,size=(len(indexes),len(columns)))

if you want to add different random values, you can make your random.randint the same size as columns

Answer 2

Create array by same size like length of indices by parameter size for 2d array:

arr = 2*np.random.randint(0,2,size=(len(indices), len(modify_columns)))
active_set.loc[indices,modify_columns] += arr

Answer 3

The easiest solution is to use pandas.DataFrame.add function.

vector_to_add = 2*np.random.randint(0, 2, size=(count)) - 1
df.loc[indices, modify_columns] = df.loc[indices, modify_columns].add(vector_to_add, axis='index')