Towards Efficient Sequential Pattern Mining in Temporal Uncertain Databases.

Uncertain sequence databases are widely used to model data with inaccurate or imprecise timestamps in many real world applications. In this paper, we use uniform distributions to model uncertain timestamps and adopt possible world semantics to interpret temporal uncertain database. We design an incremental approach to manage temporal uncertainty efficiently, which is integrated into the classic pattern-growth SPM algorithm to mine uncertain sequential patterns. Extensive experiments prove that our algorithm performs well in both efficiency and scalability.