Discovery of Characteristic Patterns from Transactions with Their Classes
Table 9
An algorithm of the refined FP growth algorithm.
(1) Extract an item from the rear of Flist in a refined FP tree . If the item is not extracted, this algorithm stops.
(2) Add it to an item set Its and investigate an identification flag of it. If the flag shows a characteristic pattern, then output as a characteristic pattern.
(3) Extract a node whose item is equal to from . If the node is not extracted, then go to step 6.
(4) Extract a path from the upper node of to the root. Generate an item set composed of item included in the extracted path and extract the frequency assigned to . Regard the combination of the item set and the frequency as a transaction conditioned by .
(5) Go to step 3.
(6) Apply a set Trans of the generated transactions to the algorithm shown in Table 6 and generate a new refined FP tree . If is not generated, then this algorithm stops. Otherwise, this algorithm recursively calls by giving and to it as arguments.
