Market basket analysis

Market basket analysis is a classic application of association rule mining. An example of association rule is

Bred -> Milk [Support = 15%, confidence = 70%]

The rule says that 15% customers buy Bred and Milk together, and those who buy Bred also buy Beer 70% of time.

The basket is considered as a transaction, which contains items. So transaction is a set of items.

For example, at a supermarket you put in the basket 1 loaf of bred and 1 carton of milk.
Transaction = {Bred, Milk}

The Support of a rule, {Bred} -> {Milk}, is

Support = countTr({Bred} + {Milk}) / N

where N is a number of transactions, countTr(Itemset) is a function that counts transations that contains Itemset.
The Confidence of a rule is

Confidence = countTr({Bred} + {Milk}) / countTr({Bred})

To prevent discovering all association rules we can use user-specified minsup and minconf.

Links