Python - Finding items that are bought together the most

Question

I have a pandas dataframe of about 40k entries in the following format:

invoiceNo | item

import pandas as pd
df = pd.DataFrame({'invoiceNo': ['123', '123', '124', '124'], 
                   'item': ['plant', 'grass', 'hammer', 'screwdriver']})

Let's say a customer can buy several items under one single invoice number.

Is there a way for me to check what items get bought together the most?

The first thing I tried was to get all unique IDs to loop through

unique_invoice_id = df.invoiceNo.unique().tolist()

Thanks!

Answer 1

Without loss of generality, I'm going to use lists instead of a dataframe. You can easily extract the lists required from the dataframe if necessary.

from itertools import combinations
from collections import defaultdict

x = [1, 1, 1, 2, 2, 2, 3, 3, 3]  # invoice number
y = ['a', 'b', 'c', 'a', 'c', 'e', 'a', 'c', 'd']  # item

z = defaultdict(set)
for i, j in zip(x, y):
    z[i].add(j)

print(z)

d = defaultdict(int)
for i in range(2, len(set(y))):
    combs = combinations(set(y), i)
    for comb in combs:
        for k, v in z.items():
            if set(comb).issubset(set(v)):
                d[tuple(comb)] += 1

list(reversed(sorted([[v, k] for k, v in d.items()])))

# [[3, ('c', 'a')],
#  [1, ('d', 'c', 'a')],
#  [1, ('d', 'c')],
#  [1, ('d', 'a')],
#  [1, ('c', 'e')],
#  [1, ('c', 'a', 'e')],
#  [1, ('b', 'c', 'a')],
#  [1, ('b', 'c')],
#  [1, ('b', 'a')],
#  [1, ('a', 'e')]]

Interpretation is 'c' and 'a' were bought together 3 times, etc.