If X and Y are independent, then the lift will be close to 1. If lift(X ? Y) > 1, then X and Y correlate positively. For example lift(X ? Y) = 5 means that X and Y happen five times more together than what would be the case if they were independent. If lift(X ? Y) < 1, then X and Y correlate negatively (i.e., the occurrence of X makes Y less likely and vice versa). Rules with a higher lift value are generally considered to be more interesting. However, typically lift values are only considered if certain thresholds with respect to support and confidence are met.
In the remainder of this section, we restrict ourselves to a special form of association rule learning known as market basket analysis. Here we only consider binary variables that should be interpreted as present or not. For example, let us consider the first two columns in Table 3.1. This data set can be rewritten to so called item-sets: