Semi-Supervised Anomaly Detection in Business Process Event Data using Self-Attention based Classification

Abstract

The analysis of business processes has become increasingly important in recent years, not least due to the emergence of analysis tools that enable data-centric views of processes and thus provide increasingly operational support for process flows. In this work, a semi-supervised classification model is presented that takes into account different developments in deep learning (e.g., deep generative models), time series analysis (e.g., long short-term memory) and sequence processing (e.g., attention mechanism) and combines them in one approach. The results of the experimental implementation of the classification model show that it is able to filter activity-related and time-related anomalies from the event data and outperform existing approaches in its classification accuracy (F1 score). The classification model achieves an F1 score of up to 93%.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 26 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.