Z-Miner: Noise filter in mining frequent access patterns

The classic mining methods do not suit to mine frequent access patterns in storage system because of their intolerable overhead. This paper analyzes noise - the major factor of low mining efficiency, and proposes a new method-Z-Miner, which employs global-branch-cutting and branch-clustering approaches to cut redundant branches in FP-Tree produced by noise. Compared to the classic methods, the overhead of Z-Miner is reduced by 1 to 2 orders of magnitude, while the effect of its directed prefetching is more than doubled.