I have used
apriori to mine roles on a training set and the result is similar to the below
I have been looking around for a kind of validation or measure of such association but it seems pretty vague and there no specific method yet, I come across an opinion which suggests using
which is defined as
TP / TP + FP
my questions are
- is using precision would be a sufficient ?
- is it correct to assume that ...
- TP =
transactions having both A and B
- FP =
transactions not containing A and B+
transcation not containig A+
transactions not containing B
- TP =
- would I just calculate the
coinefidenceon the test set ?