.png){kind=icon}
.png){kind=icon}
Return
CoTP-Miner: Co-occurrence three-way sequential pattern mining
.png){kind=icon}
PRE
AI
DOI：10.1016/j.knosys.2025.114803.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
The aim of sequential pattern mining is to extract valuable patterns from massive data. Existing methods are not targeted approaches, and the mining results may contain large numbers of patterns that users are not interested in. To address this issue, this paper proposes co-occurrence three-way sequential pattern (CoTP) mining, where the events are categorized into three levels of interest: strong, medium, and weak. Co-occurrence patterns composed of strong and medium-interest events with a given prefix pattern are mined. To mine CoTPs, an effective algorithm called CoTP-Miner is put forward, which consists of three parts: database preprocessing, support calculation, and candidate pattern generation. At the preprocessing stage, CoTP-Miner constructs a position index for all events and strong-interest events and adopts impossible sequence and non-extended sequence filtering strategies to reduce the scanning of redundant sequences. At the support calculation stage, CoTP-Miner adopts depth-first search and backtracking strategies to avoid sequential searches of the nodes at each level. At the candidate pattern generation stage, CoTP-Miner adopts a prefix-suffix pattern join method and prunes infrequent suffixes to reduce the number of candidate patterns. Experimental results on eight real datasets show that CoTP-Miner is about 2.4 to 54.7 times faster than state-of-the-art algorithms. Furthermore, compared with co-occurrence pattern mining and co-occurrence three-way pattern mining under the overlapping condition, CoTP-Miner yields a better recommendation performance in terms of the recall and F1-score.
Journal
K
IF:
7.6
Papers:
1.2W
Citations:
4.5W