Erasable itemset mining over incremental databases with weight conditions

Erasable itemset mining is an approach for mining itemsets with low profits from large-scale product databases in order to solve financial crises of plants in manufacturing industries. Previous erasable itemset mining methods deal with static product databases only, and ignore any characteristics such as items’ own values when they extract the erasable itemsets. Therefore, such approaches may fail to solve financial crises of plants because they have to iterate a significant number of mining processes in order to deal with real-time product data accumulated from plants in the real world. In this paper, we propose a new tree-based erasable itemset mining algorithm for dynamic databases, which finds erasable itemsets considering the weight conditions from incremental databases. The proposed algorithm uses new tree and list data structures for performing its mining operations more efficiently. Furthermore, the proposed algorithm is capable of reducing the number of mined erasable itemsets by considering the different weight information of items within product databases. We also compare the proposed approach with other tree-based state-of-the-art methods. By performing runtime, memory, pattern quality, and scalability comparisons with respect to various real and synthetic incremental datasets, we show that the proposed algorithm is outstanding in comparison to other previous methods.