Converting tuples to multiple indices in a Pandas Dataframe

Question

I'm starting with a dictionary like this:

dict = {(100000550L, u'ActivityA'): {'bar__sum': 14.0, 'foo__sum': 12.0},
        (100001799L, u'ActivityB'): {'bar__sum': 7.0, 'foo__sum': 3.0}}

Which, when converted to a DataFrame, puts as column headers the tuples of (id, activitytype):

df = DataFrame(dict).transpose()

                        bar__sum  foo__sum
(100000550, ActivityA)        14        12
(100001799, ActivityB)         7         3

How can I convert the tuples in the index to a MultiIndex? Ie, so that the end result looks like this instead:

                        bar__sum  foo__sum
id        act_type
100000550 ActivityA        14        12
100001799 ActivityB         7         3

What's the best way to do this? Is there some option on the DataFrame creation that I'm missing? Or should it happen via a list comprehension, which feels inefficient to me.

Answer 1

If you want to convert index of your dataframe:

>>> df.index = pd.MultiIndex.from_tuples(df.index)
>>> df
                     bar__sum  foo__sum
100000550 ActivityA        14        12
100001799 ActivityB         7         3

>>> df.index.names = ['id', 'act_type']
>>> df
                     bar__sum  foo__sum
id        act_type                     
100000550 ActivityA        14        12
100001799 ActivityB         7         3

You can also create DataFrame directly from dictionary (d is your dict, don't call your variable dict since it'll shadow standard python dictionary):

>>> pd.DataFrame(d.values(), index=pd.MultiIndex.from_tuples(d.keys(), names=['id', 'act_type']))
                     bar__sum  foo__sum
id        act_type                     
100001799 ActivityB         7         3
100000550 ActivityA        14        12

Note that values() and keys() are always in the same order, so no worries about that.