Mastering Machine Learning with Spark 2.x

Frequent pattern mining problem

Given the definition of a transaction database, a pattern P is a transaction contained in the transactions in D and the support, supp(P), of the pattern is the number of transactions for which this is true, divided or normalized by the number of transactions in D:

supp(s) = supp_D(s) = |{ s' ∈ S | s < s'}| / |D|

We use the < symbol to denote s as a subpattern of s' or, conversely, call s' a superpattern of s. Note that in the literature, you will sometimes also find a slightly different version of support that does not normalize the value. For example, the pattern {a, c, f} can be found in transactions 1, 2, and 5. This means that {a, c, f} is a pattern of support 0.6 in our database D of five items.

Support ...