specificity is defined as the probability of correctly identifying
healthy individuals. Lift was used together with confidence to
understand sensitivity and specificity.
To find predictive association rules in a medical data set the
algorithm has three major steps. First, a medical data set with
categorical and numeric attributes is transformed into a transaction
data set. Second, four constraints mentioned above are
incorporated into the search process to find predictive association
rules with medically relevant attribute combinations. Third, a train
and test approach is used to validate association rules.
Table 5 shows the 25 attributes selected for the experiments that
were the most important. The author has performed the
Experiments on a real data set to study the impact of constraints
and the elimination of unreliable rules with validation on the test
set. Figure 3 shows some of the rules with their support,
confidence and lift value.