Scholarlink

<p>High-utility itemset mining (HUIM) is an increasingly prominent data mining task focused on identifying patterns of significant importance within datasets. HUIM algorithms aim to identify all high-utility itemsets (HUIs) that meet or exceed a specified minimum utility threshold. While these algorithms can uncover a vast amount of information, not all of it is relevant to users, who are often interested only in itemsets containing specific items. Targeted mining was introduced to address this issue. However, in real-world scenarios, certain items may yield high profits only in the short term but not in the long term, which is referred to as the periodicity of itemsets. Users are generally more interested in itemsets that satisfy periodicity conditions rather than in all itemsets containing specified items. To address this need, this study proposes the problem of Targeted Mining of Periodic and High-utility Patterns and introduces an efficient algorithm, called TMPHP, for locating targeted periodic high-utility itemsets (TPHUIs). A novel pruning strategy, called partitioning pruning, is designed to enhance TMPHP’s efficiency. By utilizing this strategy, the algorithm can efficiently discard unpromising transactions, thereby reducing the number of generated candidate itemsets, lowering execution costs, and accelerating overall performance. This strategy results in approximately 10 and 13 times faster runtime on the retail and ecommerce datasets, respectively. Additionally, applying this strategy on sparse datasets reduces memory consumption by about 3 times. However, this pruning strategy is only effective on sparse datasets. To extend the effectiveness of our pruning strategy to dense datasets, we propose a new item-sorting rule. In dense datasets, the use of the new sorting rule enhances runtime by a factor of 2, reduces the number of generated candidate itemsets by 2.5 times, and does not lead to excessive memory usage. Extensive experiments demonstrate that the TMPHP algorithm is both correct and efficient. The code and datasets are publicly available on GitHub: <a href="https://github.com/DSI-Lab1/TMPHP">https://github.com/DSI-Lab1/TMPHP</a>.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Targeted mining of periodic and high-utility patterns

  • 期刊论文

  • 发表日期：

    2025年7月27日

查看全文
  • Qingfeng Zhou,
  • Wensheng Gan

摘要

High-utility itemset mining (HUIM) is an increasingly prominent data mining task focused on identifying patterns of significant importance within datasets. HUIM algorithms aim to identify all high-utility itemsets (HUIs) that meet or exceed a specified minimum utility threshold. While these algorithms can uncover a vast amount of information, not all of it is relevant to users, who are often interested only in itemsets containing specific items. Targeted mining was introduced to address this issue. However, in real-world scenarios, certain items may yield high profits only in the short term but not in the long term, which is referred to as the periodicity of itemsets. Users are generally more interested in itemsets that satisfy periodicity conditions rather than in all itemsets containing specified items. To address this need, this study proposes the problem of Targeted Mining of Periodic and High-utility Patterns and introduces an efficient algorithm, called TMPHP, for locating targeted periodic high-utility itemsets (TPHUIs). A novel pruning strategy, called partitioning pruning, is designed to enhance TMPHP’s efficiency. By utilizing this strategy, the algorithm can efficiently discard unpromising transactions, thereby reducing the number of generated candidate itemsets, lowering execution costs, and accelerating overall performance. This strategy results in approximately 10 and 13 times faster runtime on the retail and ecommerce datasets, respectively. Additionally, applying this strategy on sparse datasets reduces memory consumption by about 3 times. However, this pruning strategy is only effective on sparse datasets. To extend the effectiveness of our pruning strategy to dense datasets, we propose a new item-sorting rule. In dense datasets, the use of the new sorting rule enhances runtime by a factor of 2, reduces the number of generated candidate itemsets by 2.5 times, and does not lead to excessive memory usage. Extensive experiments demonstrate that the TMPHP algorithm is both correct and efficient. The code and datasets are publicly available on GitHub: https://github.com/DSI-Lab1/TMPHP.