In frequent itemset generation of association rule mining, what is the fundamental difference between maximal & closed patterns itemsets. Can someone guide me a resource about them?
A closed pattern is a frequent pattern. So it meets the minimum support criteria. In addition to that, all super-patterns of a closed pattern are less frequent than the closed pattern.
Let’s see some examples.
Suppose, the minimum support count is 2. For the first example, suppose there are a total of 3 items: a, b, c. Suppose a pattern ab has support count of 2 and a pattern abc has support count of 2. Is the pattern ab is a closed pattern? Pattern ab is a frequent pattern, but it has a super-pattern that is NOT less frequent than ab.
For the second example,
suppose there are a total of 3 items: x, y, z. suppose a pattern xy has support count of 3 and a pattern xyz has support count of 2. Is the pattern xy is a closed pattern? Pattern xy is a frequent pattern and also the only super-pattern xyz is less frequent than xy.
Therefore, xy is a closed pattern.
A max pattern is
a frequent pattern. So it also meets the minimum support criteria like closed pattern In addition, but unlike closed pattern, all super-patterns of a max pattern are NOT frequent patterns.
Let’s see some examples as well.
Suppose, the minimum support count is 2. Like before, for the first example, suppose there are a total of 3 items: a, b, c. Suppose a pattern ab has support count of 3 and a pattern abc has support count of 2. Is the pattern ab is a max pattern? Pattern ab is a frequent pattern, but it has a super-pattern that is a frequent pattern as well. So, pattern ab is NOT a max pattern.
For the second example,
suppose there are a total of 3 items: x, y, z. Suppose a pattern xy has support count of 3 and a pattern xyz has support count of 1. Is the pattern xy is a max pattern? Pattern xy is a frequent pattern and also the only super-pattern xyz is NOT a frequent pattern. Therefore, xy is a max pattern.