HDminer: Efficient Mining of High Dimensional Frequent Closed Patterns from Dense Data

Frequent closed pattern mining has been developed for decades, mostly on a two dimensional matrix. This paper addresses the problem of mining high dimensional frequent closed patterns (nFCPs) from dense binary dataset, where the dataset is represented by a high dimensional cube. As existing FP-tree or enumeration tree based algorithms do not suit for n-dimensional dense data, we are motivated to propose a novel algorithm called HDminer for nFCPs mining. HDminer employs effective search space partition and pruning strategies to enhance the mining efficiency. We have implemented HDminer, and the performance studies on synthetic data and real microarray data show its superiority over existing algorithms.