An Enhanced Frequent Pattern-Growth Algorithm with Dual Pruning using Modified Anti-Monotone Support

Pattern discovery does not only end when a process obtained a certain pattern. It also requires careful evaluation to show whether the pattern is significant enough to support any decision-making. Generating interesting frequent pattern is important to remove uninteresting and weak rules. The study, Dual Pruned Frequent Pattern-Growth (2P FP-Growth), enhanced the FP-Growth algorithm by performing the dual pruning of the itemsets before generating frequent patterns. The 2P FP-Growth algorithm first removed the itemsets not satisfying the minimum support count, which represent the first pruning. Consequently, the algorithm constructed the FP tree. Secondly, the 2P FP-Growth traverses each subtree, removing the nodes in a subtree that do not satisfy the minimum support count, constituting the second pruning. The second pruning process used the modified anti-monotone support constraint which removes the nodes in a subtree that do not meet with the minimum support count, but not the entire subtree. The confidence measure of the resulting frequent patterns of the 2P FP-Growth showed the most interesting frequent patterns with a confidence measure of 1.0, while obtaining the interdependency result of less than 1.0. The comparative result implies that the association rule's performance is negatively interdependent to its predicted response.