Alternative Method for Increnentally Constructing the FP-Tree

The FP-tree is an effective data structure that facilitates the mining of frequent patterns from transactional databases. But, transactional databases are dynamic in general, and hence modifications on the database must be reflecting onto the FP-tree. Constructing the FP-tree from scratch and incrementally updating the FP-tree are two possible choices. However, from scratch construction turns unfeasible as the database size increases. So, this paper addresses incremental update by extending the FP-tree concepts and manipulation process. Our new approach is capable of handling all kinds of changes; include additions, deletions and modifications. The target is achieved by constructing and incrementally dealing with the complete FP-tree, i.e., with one minimum support. Constructing the complete FP-tree has the other advantage that it provides the freedom of mining for lower minimum support values without the need to reconstruct the tree. However, directly reflecting the changes onto the FP-tree may invalidate the basic FP-tree structure. Thus, we apply a sequence of shuffling and merging operations to validate and maintain the modified tree. The experiments conducted on synthetic and real datasets clearly highlight advantages of the proposed incremental approach over constructing the FP-tree from scratch