Efficient high utility itemsets mining over data streams with compact utility list structure

Efficient high utility itemsets mining over data streams with compact utility list structure

연구 분야: Databases

학회: Applied Intelligence

With the growing prevalence of real-time data in various application domains, data streams have become an important focus in the field of data mining. High utility itemset mining (HUIM) over data streams is particularly challenging, as it must ensure both efficiency and accuracy in a continuously evolving environment. To address these challenges, this paper proposes a novel algorithm named CHUPM-Stream, which leverages Compact Utility Lists (CU-lists) to mine high utility itemsets within a sliding window model. The algorithm effectively captures transaction utility information using the CU-list structure and employs a Duplicate Transaction Merge (DTM) strategy to compress redundant transactions, thereby reducing memory consumption. To avoid repeated scanning of the entire window, CHUPM-Stream further introduces a Transaction weighted utility Maintenance Strategy (TMS), which reuses TWU values from the previous window. Additionally, it applies the LA-prune strategy, taking advantage of CU-list characteristics to reduce the search space. Experimental results on dense datasets demonstrate that CHUPM-Stream achieves superior performance compared to state-of-the-art algorithms in terms of both runtime and memory usage.