Answers are grounded in this paper and cite their sources.
Computational evaluation demonstrates superior efficiency of bidirectional pattern mining in sequential datasets, highlighting scalable discovery of closed contiguous patterns.
Sequential pattern mining is a cornerstone of data-driven discovery, yet existing methods often struggle with the dual constraints of contiguity and closure–essential for preserving the semantic integrity of sequences in domains such as genomics, sensor networks, and linguistics. This paper introduces TPECSM (Targeted Pattern Extension for Contiguous Sequences Mining), a novel algorithm specifically engineered for the efficient discovery of closed contiguous patterns. Unlike traditional unidirectional expansion methods, TPECSM employs a novel bidirectional extension strategy that explores the search space on both the right and left sides of a query motif. We provide a mathematical framework to prove that this strategy ensures both completeness and the elimination of redundancy without the computational overhead of candidate maintenance. Our experimental evaluation, conducted on diverse real-world datasets including retail and e-commerce behavior, demonstrates that TPECSM significantly outperforms state-of-the-art methods, particularly when mining infrequent patterns where traditional algorithms become computationally prohibitive. By reducing the output size through a closed-search strategy while maintaining full informative value, TPECSM provides a scalable solution for extracting high-utility insights from large-scale ordered data. These findings offer a robust computational tool for researchers requiring precise, non-redundant pattern recognition in complex sequential environments.
No takes yet. Share an insight, caveat, or question.
Pasquier et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: