OHUQI: Mining on-shelf high-utility quantitative itemsets

Mobile edge computing has brought fresh opportunities and challenges to data science. Utility-driven mining, a recently emerging branch of utility-based data science, has been widely applied because it considers both the utility factor and the quantity characteristic with ranges of patterns. However, most existing utility-mining algorithms assume that patterns always appear regardless of the period. For instance, some products may sell well at certain times of the year. Considering the rich information in the database, such as quantity and time, we propose an effective and efficient approach, namely OHUQI, for discovering on-shelf high-utility quantitative itemsets. To avoid scanning the database multiple times, we adopt a data structure to maintain some necessary information, and thus OHUQI only accesses the database twice. Several pruning strategies are also designed to prune a large number of unpromising itemsets in advance to shrink the search space. Finally, the subsequent experimental results show that OHUQI performs well on several real-world datasets.