ValueError: could not convert string to float: '62,6'

Question

I am trying to convert dataframe to numpy array:

dataset = myset.values
X = np.array(dataset[0:,6:68], dtype="float32")
X[0:5,0:]

Here is a piece of the data

Here is an error:

-----------------------------------------------------------------------

----
ValueError                                Traceback (most recent call last)
<ipython-input-162-4b67608047d1> in <module>()
      1 dataset = myset.values
----> 2 X = np.array(dataset[0:,6:68], dtype="float32")
      3 X[0:5,0:]

ValueError: could not convert string to float: '62,6'

Where is a problem?

Answer 1

Try use replace , to .:

dataset = myset.replace(',','.', regex=True).values

Or use parameter decimal in read_csv for convert , to . in floats:

dataset = pd.read_csv('file', decimal=',')

Answer 2

There are locales that use '.' as decimal separator and ',' as thousands-separator. There are also locales that use ',' as decimal separator and '.' as thousands-separator.

Doing a plain string-replace will get you in all sorts of trouble. You need to

import locale

then specify which locale you want to use, e.g.

locale.setlocale(locale.LC_NUMERIC, 'german')

and then you can parse the string using

locale.atof('1.337,3')