Correlated time-window constrained high-utility itemsets mining with certain and uncertain real-life datasets

High-utility Itemsets mining (HUIM) plays a crucial role in extracting valuable patterns from transaction data, aiding decision-making across industries. However, existing approaches often generate overwhelming numbers of patterns, making it difficult to extract actionable insights. This paper introduces the Correlated Time-Window Constrained High-Utility Itemsets Mining (CTC-HUIM) algorithm, addressing this challenge. CTC-HUIM significantly reduces processing time compared to existing methods like FFU-TSPM, as demonstrated in our experiments. The core novelty lies in the introduction of two novel pruning strategies T-HUDS and top-k revised transaction maximum utility upper-bound (KRTMUUB). These strategies effectively limit the candidate set size, leading to a more focused search space and faster execution. CTC-HUIM further incorporates time-window constraints, enabling the discovery of relevant patterns within specific timeframes. This is particularly valuable for applications dealing with dynamic data. The effectiveness of CTC-HUIM is validated using both certain and uncertain real-life datasets. The proposed approach demonstrates superior performance in terms of runtime, memory usage, and pattern discovery compared to existing methods. Notably, CTC-HUIM achieves a runtime of only 215.23 s, highlighting its efficiency. In conclusion, CTC-HUIM offers a significant advancement in HUIM by reducing processing time, improving efficiency, and enabling time-constrained pattern discovery. This paves the way for extracting more insightful patterns from high-volume transaction data, empowering better decision-making in real-world scenarios.