區間型事件序列樣式探勘

Mining Temporal Patterns from Interval-Based Event Sequences

吳欣怡(Shin-Yi Wu)

《電腦與通訊》 123期 (2008/03) Pp. 130-137

https://doi.org/10.29917/CCLTJ.200803.0019

摘要

序列樣式探勘可應用顧客行為分析、經濟指標分析、財稅預測、詐欺偵測等領域，以輔助決策制定或針對重要事件進行預測。然而，過去的序列樣式探勘技術多為針對單點型事件所設計，也就是說，這些技術所探勘的序列資料中，所有的事件皆發生於某個時間點。然而，在許多應用中，事件不盡然皆發生於某個時間點，而可能持續發生於一段時間，這樣的事件稱為區間型事件。在區間型事件所組成的序列中尋找頻繁樣式，稱為區間型序列樣式探勘。本文除了介紹此一新探勘問題外，亦提供數個潛在應用情境，此外，本文更進一步介紹及討論目前針對此一問題所發展的各種探勘方法，期望為此項新技術帶來更多應用的契機。

關鍵字

資料探勘 ； 序列樣式 ； 時間樣式 ； 區間型事件序列

並列摘要

Sequential pattern mining is useful in various domains, such as customer behavior analysis, economic indices analysis, tax prediction, and fraud detection. Discovering sequential patterns helps make decisions or predict key events. Previous studies of sequential pattern mining have discovered patterns from point-based event sequences. However, in some applications, event sequences may contain interval-based events. Frequent patterns discovered from interval-based event sequences are called temporal patterns. This paper explains the new mining problem-discovering temporal patterns from interval-based event sequences. Besides, we describe several scenarios which are the possible domain the temporal pattern mining methods could be applied to. Finally, we introduce the existing temporal pattern mining methods and point out their problems. We hope this new mining technique could be applied to more and more applications.

並列關鍵字

Data Mining ； Sequential Patterns ； Temporal Patterns ； Interval-based Event Sequence