Since I haven't found a good-enough package in Python for association rules learning, I am manually coding up my own.
However, I have seen conflicting information about exactly what the formulae for support, confidence, and lift (three association rules) are. For example, one resource tells me that, when looking at the support of two items A and B in the transaction basket T, you're supposed to examine all cases when A and B occur in T and divide that by the number of transactions. Another resource tells me it's A or B. Can anyone share the canonical formulae for these rules?