P-FCloHUS: A Parallel Approach for Mining Frequent Closed High-Utility Sequences on Multi-core Processors

Frequent closed high-utility (FCHU) sequences are preferable to frequent closed sequences. Not only because of their utility-based nature that considerately contributes to taking decisive business actions, FCHU sequences also preserve necessary information for re-constructing frequent high-utility sequences. Despite of their vital role, mining FCHU sequences is a time consuming task when facing with large-scale datasets, or especially when the input thresholds are relatively small. To contend with these difficulties, this paper proposes a parallel algorithm named P-FCloHUS for fast mining FCHU sequences by making good use of multi-core processors. By relying on a novel Single scan synchronization strategy that is facilitated by an efficiently Partitioned result space structure, P-FCloHUS successfully alleviates the communication cost between mining tasks and hence speeds up the parallel mining process. Experiments on both dense and sparse datasets show that P-FCloHUS outperforms the state-of-the-art FMaxCloHUSM in terms of runtime performance.