Question

Consider two association rules mined from some shopping basket dataset: “cereal ⇒ milk” and “milk ⇒ cereal”. Which of the following is true about the lift of these rules? Explain.

1. Lift(cereal ⇒ milk) = Lift(milk ⇒ cereal).
2. Lift(cereal ⇒ milk) ≠ Lift(milk ⇒ cereal).
3. Not enough information, i.e., it can be either of the above, depending on the specific dataset.

Answer 1

lift “cereal ⇒ milk” = (Number of transactions with cereal and milk) / (Number of transactions with milk) * (Number of transactions with cereal)

or Support (“cereal ⇒ milk” ) / {Support(cereal) * Support (Milk) }

Now, the numerator and denominator will remain the same for support and lift values of both the rules, “cereal ⇒ milk” and “milk ⇒ cereal”. Hence,

1. Lift(cereal ⇒ milk) = Lift(milk ⇒ cereal).