Convert specific value in numpy array to NAN

Question

I have a two-dimensional numpy-array. It has a shape of (6994, 6994). There are many values of -1000 which I would like to encode as NAN. I tried:

array[array == -1000] = np.NAN, but this gives me the error cannot convert float NaN to integer

When I tried to write a function:

def valtona(array, val):
    for i in array:
        for j in array:
            if array[i,j] == -1000:
                array[i,j] = np.NAN

I get: ValueError: The truth value of an array with more than one element is ambiguous. Use a.any() or a.all()

I know there are some questions out there regarding the same issue, but I still didn't manage to fix it.

np.nan is a float. Your array is integer. For the sklearn use which is more useful? — hpaulj
– hpaulj, Commented Feb 29, 2020 at 16:22

0buz · Accepted Answer · 2020-02-29 14:51:34Z

4

You can still use

array[array == -1000] = np.NAN

You just need to convert it to float first.

array=array.astype('float')
array[array == -1000] = np.NAN

answered Feb 29, 2020 at 14:51

0buz

3,5352 gold badges12 silver badges31 bronze badges

Sign up to request clarification or add additional context in comments.

4 Comments

Guimoute Over a year ago

Enjoy the comparison issues where float 1000.000000000001 != int 1000.

0buz Over a year ago

Fair point, however the conversion will be to float64, not float32, with significantly better precision.

Robin Kohrs Over a year ago

perfect, helped a lot!!

hpaulj Over a year ago

The mask could be calculated with the original int dtype array.

Ethan · Accepted Answer · 2020-02-29 16:19:35Z

1

You can use np.isclose() and set parameters to meet your needs to overcome the precision challenge of working with floats.

>>> a
array([ 0.,  1.,  2.,  4.,  4.,  5.,  6.,  7.,  8.,  9.])
>>> a[3]
4.0000000000001004
>>> a[4]
4.0
>>> np.isclose(a,[4.0], .00000001, .00000001)
array([False, False, False,  True,  True, False, False, False, False, False], dtype=bool)
>>> np.isclose(a,[4.0])
array([False, False, False,  True,  True, False, False, False, False, False], dtype=bool)
>>> a[np.isclose(a,[4.0], .00000001, .00000001)]=np.nan
>>> a
array([  0.,   1.,   2.,  nan,  nan,   5.,   6.,   7.,   8.,   9.])

answered Feb 29, 2020 at 16:19

Ethan

1,3831 gold badge8 silver badges8 bronze badges

Collectives™ on Stack Overflow

Convert specific value in numpy array to NAN

2 Answers 2

4 Comments

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

4 Comments

Comments

Your Answer

Sign up or log in

Post as a guest

Related