A Novel Algorithm for Mining Closed Sequential Patterns

Abstract

Sequential pattern mining algorithms produce an exponential number of sequential patterns when mining long patterns or at low support thresholds. Most of the existing algorithms mine the full set of sequential patterns. However, it is sufficient to mine closed sequential patterns from which the total set of sequential patterns can be derived and the closed sequential patterns set is more compact than the sequential patterns set. In this paper, we propose a novel algorithm NCSP for mining closed sequential patterns in large sequences databases. To the best of our knowledge, our algorithm is the first algorithm that utilizes vertical bitmap representation for closed sequential pattern mining. The results show that the proposed algorithm NCSP can find closed sequential patterns efficiently and outperforms CloSpan by an order of magnitude.