High utility itemset mining in data stream using elephant herding optimization

Mining high utility itemsets from data stream within limited time and space is a challenging task. Traditional algorithms typically require multiple scans and complex data structures for data connection, storage and update. Moreover, the evaluation of duplicate itemsets generated by overlapping batches leads to low efficiency of the algorithm in terms of time and space. To address these issues, this paper proposes a heuristic-based data stream high utility itemset mining algorithm, termed SHUIM-EHO, designed to effectively solve limited storage space. The SHUIM-EHO algorithm designs a new clan updating strategy, which effectively enhances the convergence speed and reduces itemset loss. Additionally, a hash storage strategy is proposed to avoid the evaluation of duplicate itemsets, thereby further improving the execution efficiency of the algorithm. Experiments on real and synthetic datasets demonstrate the effectiveness of the algorithm, significantly reducing memory consumption and maintaining strong scalability.