Association rules

Question 1

Support of a rule

Answer 1

How often do the LHS and RHS appear together

Question 2

What is confidence

Answer 2

Given LHS, how often do we see RHS

Question 3

What is lift

Answer 3

How often does LHS appear with RHS, compared to what chance would predict?

Question 4

Confidence (X->Y)

Answer 4

n(X & Y) / n(X)

Question 5

Is confidence symmetric

Answer 5

No

Question 6

Lift formula (X->Y)

Answer 6

c(X->Y)/s(Y)

s(X&Y)/s(X) x s(Y)

Question 7

Interpret lift (X->Y) = 5

Answer 7

Shoppers who buy X are 5x (500%) more likely to buy Y than the average shopper

Question 8

Lift = 1 interpretation

Answer 8

Indicates almost no correlation

Events seem independent

Question 9

Lift > 1 interpretation

Answer 9

Positive correlation
Cooccurrence is more likely than chance
Complementary products that are often bought together

Question 10

Lift <1

Answer 10

Indicates negative correlation - co-occurance is less likely than chance
Customers who buy one item (or a given item set) are less likely (than average) to purchase the other

Substitutes