Damped weighted erasable itemset mining with time sensitive dynamic environments

Erasable itemset mining discovers itemsets in product databases with benefits no greater than a designated threshold value. By considering weight constraints and the recency of products in erasable itemset mining, the practitioners can manage the plants more efficiently. However, existing studies on weighted erasable itemset mining do not regard the time-sensitivity of data arrival times. In this paper, we propose a new weighted erasable itemset mining approach considering time-sensitive dynamic environments. For industrial manufacturers with automated control systems, our method focuses on recent data and item weights to effectively filter out unprofitable itemsets. Performance tests show that the proposed method outperforms state-of-the-art studies in runtime with on-par or minimal compromise in memory usage and that it scales capably with varying database sizes. We performed statistical analyses to demonstrate the correctness and significance of the discovered results from our proposed method. Furthermore, extended evaluations on sensitivity and resultant itemsets show how the algorithm responds to varying parameters as well as provide insights on the discovered itemsets.