High-Utility Sequential Pattern Mining with Multiple Minimum Utility Thresholds

High-utility sequential pattern mining is an emerging topic in recent decades and most algorithms were designed to identify the complete set of high-utility sequential patterns under the single minimum utility threshold. In this paper, we first propose a novel framework called high-utility sequential pattern mining with multiple minimum utility thresholds to mine high utility sequential patterns. A high-utility sequential pattern with multiple minimum utility thresholds algorithm, a lexicographic sequence (LS)-tree, and the utility-linked (UL)-list structure are respectively designed to efficiently mine the HUSPs. Three pruning strategies are then introduced to lower the upper-bound values of the candidate sequences, and reduce the search space by early pruning the unpromising candidates. Substantial experiments on real-life datasets show that our proposed algorithms can effectively and efficiently mine the complete set of HUSPs with multiple minimum utility thresholds.