Prefix Imputation of Orphan Events in Event Stream Processing

Area of Science:

• Process Mining
• Data Science
• Information Systems

Background:

• Online conformance checking processes streaming event logs to monitor process adherence.
• Existing methods face memory limitations and can misinterpret 'orphan events' (events from forgotten cases).
• This can lead to inaccurate conformance statistics and misleading process insights.

Purpose of the Study:

• To propose a novel approach for handling orphan events in online conformance checking.
• To bound memory usage effectively while maintaining accurate process insights.
• To improve the realism of conformance statistics in streaming environments.

Main Methods:

• Developed an imputation method for orphan events using the process model to reconstruct missing prefixes.
• Implemented a case storage management strategy to enhance prefix prediction accuracy.
• Introduced a systematic forgetting mechanism for cases that can be regenerated.

Main Results:

• The proposed imputation approach effectively incorporates orphan events without unbounded memory.
• Achieved considerably higher realistic conformance statistics compared to state-of-the-art methods.
• Maintained the same storage requirements as existing techniques.

Conclusions:

• The imputation of missing prefixes offers a robust solution for managing orphan events in online conformance checking.
• This method enhances the accuracy and reliability of conformance checking in streaming data.
• It provides a practical advancement for real-time process monitoring and analysis.