Open Access

ARTICLE

Mining High-Quantitative Periodic Frequent Patterns across Multiple Sequences

Yan Ge¹, Zhenzhou Zhang², Chien-Ming Chen^3,*

1 Reading Academy, Nanjing University of Information Science and Technology, Nanjing, China
2 China Unicom (Shandong) Industrial Internet Co., Ltd., Jinan, China
3 School of Artificial Intelligence, Nanjing University of Information Science and Technology, Nanjing, China

* Corresponding Author: Chien-Ming Chen. Email:

Computers, Materials & Continua 2026, 88(2), 38 https://doi.org/10.32604/cmc.2026.077790

Received 17 December 2025; Accepted 17 April 2026; Issue published 15 June 2026

Abstract

Periodic pattern mining plays an important role in revealing recurring behavioral regularities from temporal sequence data. Most existing approaches, however, are developed for single-sequence settings and rarely account for quantitative information or sequence-level constraints when patterns recur across multiple sequences. This limits their usefulness in practical scenarios, where a pattern is expected to be not only periodic but also quantitatively significant in a sufficiently large portion of sequences. In this work, we formulate the problem of mining High-Quantitative Periodic Frequent Patterns (HQPFPS) from multi-sequence databases and propose an efficient algorithm, termed MHQPFPS. The proposed method evaluates pattern significance through a quantitative ratio within each sequence and exploits a sequence-level upper bound to effectively prune unpromising candidates during pattern growth. To support efficient evaluation, a compact list-based structure is introduced to maintain support, periodicity, and quantitative statistics, thereby avoiding repeated scans of the database. These components are combined within a depth-first exploration framework to systematically generate valid patterns while discarding those that fail to satisfy the required periodic or quantitative constraints. Experimental results on both real-world and synthetic datasets show that MHQPFPS is able to extract meaningful high-quantitative periodic patterns across multiple sequences. Moreover, the results indicate that the proposed pruning strategies substantially reduce computational cost in terms of runtime and memory consumption under a wide range of parameter settings.

---

Keywords

---