APT-Structure: Efficient Mining of Frequent Patterns

Frequent pattern mining is a key step in many data mining applications. In this paper, we propose a simple and novel pattern growth algorithm, which uses a compact data structure named Array-based Prefix Tree (APT). The APT has a distinct feature that the space requirement can be predictable in advance. The memory usage of APT is less than FP-Tree that uses pointer to maintain the link between parent and child nodes, and the traversal cost is lower. The mining algorithm based on APT uses top-down traversal strategy, and unfiltered pseudo-construct conditional database, which can improve computational performance. Further computational experiments show that APT algorithm is more efficient, and performs better than FPGrowth* and AFOPT.