The Concept of Lift in Association Rules Mining

Let's look at a problem first:


    

        

             
        

        

            Coffee
        

        

            Not Coffee
        
        

             
        
    

    

        

            Tea
        

        

            150
        

        

            50
        

        

            200
        
    

    

        

            Not Tea
        

        

            650
        

        

            150
        

        

            800
        
    

    

        

             
        

        

            800
        

        

            200
        

        

            1000
        
    


Tea -> Coffee
What is the support and confidence of the above rule?

Support = #(tea and coffee) / #(total)

Confidence : P (Coffee | Tea)
= #(tea and coffee) / #(tea)

Support for the rule is 15% while the confidence is 75%.

Lift = Confidence(A -> B) / Support(B)
= Confidence of the rule / Support of the consequent

If you look in general:
The percentage of people drinking coffee is 80%.

But when we apply the rule: Tea -> Coffee
Confidence: Percentage of people drinking coffee who are also drinking tea is:
150 / 200 = 75%

So, ( Tea -> Coffee ) shows they are negatively correlated.
This is the importance of lift.

Formula 1: Lift = P (A and B) / ( P(A) * P(B) )

Formula 2: Lift = Ratio between the rule's confidence and support of the rule consequent = c(A -> B) / s(B)

Definition

Life is a measure of the performance of an association rule at predicting or classifying cases as having an enhanced reponse (with respect to the population as a whole), measured against a random choice targeting model. A targeting model is doing a good job if the response within the target is much better than the average for the population as a whole.

Tags: Technology,Machine Learning