Closed Structured Patterns and Motifs Mining without Candidate Maintenance

Abstract

Structured motif is a special kind of frequent sequential pattern, which consists of several components separated by gaps, with important applications, especially in DNA sequence analysis. Mining closed structured patterns can get a more compact and complete result set by eliminating redundant patterns subsumed by other super patterns. However, the traditional method, maintaining candidates and testing which ones are closed, is inherently costly in both runtime and space usage. In this paper, we present BMCM, an efficient algorithm for mining closed structured patterns and motifs. It adopts BI-Composite scheme to generate, prune patterns and check patterns' closure, in which do not need maintain candidates. The experimental evaluation with synthetic data and biological data demonstrates the algorithm BMCM is effective in mining closed structured patterns and motifs.