Recent prelarge-based research to explore utility streams has gained attention for efficient resource usage by minimizing resource-intensive pattern expansion procedures by leveraging previously discovered results. However, in recency-based scenarios where more recent data carry greater importance, since these methods treat all data equally, their analysis yields outdated trends. Although some approaches consider temporal importance, they must perform the costly pattern expansion procedure upon each data arrival due to the lack of a prelarge concept, resulting in high computational cost. To overcome this limitation, we suggest a novel method adopting the prelarge concept while exploring utility streams with a time-fading effect. The proposed method efficiently assigns greater importance to more recent transactions within a list-based data structure based on a time-fading effect. Moreover, our method utilizes previously discovered patterns to detect results without performing costly pattern expansions when a re-scan condition is not met upon the arrival of new stream data. Extensive performance evaluations demonstrate that our approach is more efficient than prior methods. It achieves enhanced runtime, up to 2.9 times faster than the most efficient state-of-the-art approach, and stable scalability, while maintaining comparable memory usage without any result loss with 100% accuracy. Statistical hypothesis testing shows that the proposed method detects approximately four times fewer results on average than the prior study. The qualitative analysis shows that the proposed method extracts a more refined and recency-focused results set than the up-to-date study, while the case study demonstrates its practical applicability. • Reducing costly pattern expansions to explore recency-focused utility streams. • Maintaining prior results efficiently with time-fading effect without any loss. • Refining the re-scan condition under time-fading effect to prevent any loss. • Exhibiting enhanced runtime and scalability in diverse utility stream scenarios.
Kim et al. (Wed,) studied this question.