Tell me more ×
Stack Overflow is a question and answer site for professional and enthusiast programmers. It's 100% free, no registration required.
up vote 3 down vote favorite
2

I'm trying to convert a string array of categorical variables to an integer array of categorical variables.

Ex.

import numpy as np
a = np.array( ['a', 'b', 'c', 'a', 'b', 'c'])
print a.dtype
>>> |S1

b = np.unique(a)
print b
>>>  ['a' 'b' 'c']

c = a.desired_function(b)
print c, c.dtype
>>> [1,2,3,1,2,3] int32

I realize this can be done with a loop but I imagine there is an easier way. Thanks.

share|improve this question

3 Answers

up vote 2 down vote accepted

Well, this is a hack... but does it help?

In [72]: c=(a.view(np.ubyte)-96).astype('int32')

In [73]: print(c,c.dtype)
(array([1, 2, 3, 1, 2, 3]), dtype('int32'))
share|improve this answer
up vote 3 down vote

np.unique has some optional returns

return_inverse gives the integer encoding, which I use very often

>>> b, c = np.unique(a, return_inverse=True)
>>> b
array(['a', 'b', 'c'], 
      dtype='|S1')
>>> c
array([0, 1, 2, 0, 1, 2])
>>> c+1
array([1, 2, 3, 1, 2, 3])

it can be used to recreate the original array from uniques

>>> b[c]
array(['a', 'b', 'c', 'a', 'b', 'c'], 
      dtype='|S1')
>>> (b[c] == a).all()
True
share|improve this answer
up vote 3 down vote

One way is to use the categorical function from scikits.statsmodels. For example:

In [60]: from scikits.statsmodels.tools import categorical

In [61]: a = np.array( ['a', 'b', 'c', 'a', 'b', 'c'])

In [62]: b = categorical(a, drop=True)

In [63]: b.argmax(1)
Out[63]: array([0, 1, 2, 0, 1, 2])

The return value from categorical (b) is actually a design matrix, hence the call to argmax above to get it close to your desired format.

In [64]: b
Out[64]:
array([[ 1.,  0.,  0.],
       [ 0.,  1.,  0.],
       [ 0.,  0.,  1.],
       [ 1.,  0.,  0.],
       [ 0.,  1.,  0.],
       [ 0.,  0.,  1.]])
share|improve this answer
Neat and clever. Thanks. – unutbu Jul 10 '10 at 11:34

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.