fpgrowth is creating dataframe column 'antecedent' of type str. Why is type not frozenset?

[geoffreya]

It is the case that frozenset (an actual frozenset) is the type of the dataframe column being created by a unit test in https://github.com/rasbt/mlxtend/blob/master/mlxtend/frequent_patterns/tests/test_fpbase.py

However, str is the type in the dataframe column 'antecedent' being created by fpgrowth, not an actual frozenset. It is a actual string that looks like this example:

"frozenset({'CollegeBasketball', 'nfl'})"

For now, we have to first run eval() on such an antecedent str, in order to get an antecedent that is actually of type frozenset, for example:

eval("frozenset({'CollegeBasketball', 'nfl'})")

Instead, like the unit test, perhaps frozenset should also be the type that fpgrowth is creating.

What is your opinion?

Thank you

[rasbt]

Thanks for the comment. I am not sure if I am on the same page. (Haven't looked at the frequent pattern code in a while).

So, fpgrowth should be returning frozensets via the dataframe column entries:

from mlxtend.frequent_patterns import fpgrowth
import pandas as pd
from mlxtend.preprocessing import TransactionEncoder


dataset = [['Milk', 'Onion', 'Nutmeg', 'Kidney Beans', 'Eggs', 'Yogurt'],
           ['Dill', 'Onion', 'Nutmeg', 'Kidney Beans', 'Eggs', 'Yogurt'],
           ['Milk', 'Apple', 'Kidney Beans', 'Eggs'],
           ['Milk', 'Unicorn', 'Corn', 'Kidney Beans', 'Yogurt'],
           ['Corn', 'Onion', 'Onion', 'Kidney Beans', 'Ice cream', 'Eggs']]


te = TransactionEncoder()
te_ary = te.fit(dataset).transform(dataset)
df = pd.DataFrame(te_ary, columns=te.columns_)

df = fpgrowth(df, min_support=0.6, use_colnames=True)

>>> type(df['itemsets'][0])
frozenset

The dataset columns are "support" and "itemset"

So, I was wondering if you were perhaps referring to the association_rules function? However, also here, it seems like it is already a frozenset:

from mlxtend.frequent_patterns import association_rules

df_2 = association_rules(df, metric="confidence", min_threshold=0.7)

>>> type(df_2['antecedents'][0])
frozenset

So, I am wondering if there is a bug in your case maybe, or are you referring to how something is handled in the code internally (I haven't looked at the details in a while and would have to refresh my memory then)