Loading Arff Data not working about scikit-multilearn

[niedakh]

@ChristianSch thank you for the fix!

[ChristianSch]

Hey there,
the docs are wrong, you can load arff data sets like this:

from skmultilearn.dataset import load_dataset
d = load_dataset('enron', 'undivided')

d is a dictionary which (in the case of enron) gives you X, y or features and labels as keys:

X, y = d['X'], d['y']

Available data sets can be listed like this:

from skmultilearn.dataset import available_data_sets
available_data_sets()

For more information please see here.

[mattalhonte]

Not working here either, on Python 3.5 with Anaconda. First time, it failed with trying to load the Standard Library from future. Monkeypatched that import out with a comment.

Now I'm getting this:
from skmultilearn.dataset import available_data_sets
available_data_sets()

TypeError                                 Traceback (most recent call last)
<ipython-input-73-c3b496b1614f> in <module>()
      1 from skmultilearn.dataset import available_data_sets
----> 2 available_data_sets()

~/miniconda3/envs/py35/lib/python3.5/site-packages/skmultilearn/dataset.py in available_data_sets()
    106     archives = get_dataset_list()
    107     archives = [x.split(';')[-1].split('.')[0].split('-') for x in archives.split('\n')]
--> 108     variants = set()
    109     for a in archives:
    110         if a[0] not in variants:

TypeError: a bytes-like object is required, not 'str'

[ljvmiranda921]

I think this is the same bytes/string issue regarding the incompatibility between Python 2 and 3.
To the creators, I'm wondering if you are doing an integration test (like with travis-ci) to atleast check the compatibility.

[ChristianSch]

I can replicate this problem for a clean install of python 3.6, however not for python 2. I'll look into it.