itemset mining with utility using Pandas DataFrames

[aymenbensaid1]

is it possible to use two pandas dataframes for the transactions and the external utilities? Thank you very much in advance.

[ZaxR]

Even though this was closed by the OP, I'm going to answer this for others who may be interested. Currently there is no pandas support. The structure that would likely make the most sense for transaction data in a dataframe would be "long form," with a row for every transaction-item-internal utility. Taking the data from the example in the readme:

transactions = [
    [("Coke 12oz", 6), ("Chips", 2), ("Dip", 1)],
    [("Coke 12oz", 1)],
    [("Coke 12oz", 2), ("Chips", 1)],
    [("Chips", 1)],
    [("Chips", 2)],
    [("Coke 12oz", 6), ("Chips", 1)]
]

and converting it into the described pandas format for example purposes:

import pandas as pd

rows = []
for t_id, t in enumerate(transactions):
    for item_tuple in t:
        rows.append([t_id, *item_tuple])
transactions_df = pd.DataFrame(rows, columns=["transaction_id", "item", "internal_utility"])

would get you something like:

   transaction_id       item  internal_utility
0               0  Coke 12oz                 6
1               0      Chips                 2
2               0        Dip                 1
3               1  Coke 12oz                 1
4               2  Coke 12oz                 2
5               2      Chips                 1
6               3      Chips                 1
7               4      Chips                 2
8               5  Coke 12oz                 6
9               5      Chips                 1

To use this library as is, you could do the following conversion:

transactions = transactions_df.groupby("transaction_id")[
    ["item", "internal_utility"]
].apply(lambda x: tuple(x.values))