Anomaly Correction of Business Processes Using Transformer Autoencoder

TL;DR

This work tackles the dual problem of detecting and correcting anomalies in business process event logs. It introduces TransformerAE, a Transformer-based encoder–decoder that reframes anomaly detection as a classification task and performs anomaly correction in a single forward pass, guided by a joint loss $L = L_{detect} + L_{reconstruct}$ with $L_{detect}$ as binary cross-entropy and $L_{reconstruct}$ as sequence cross-entropy. By injecting synthetic anomalies to train the model and using a self-supervised framework, TransformerAE achieves strong performance on four real BPIC datasets, outperforming baselines in both detection accuracy and correction quality while maintaining high efficiency. The approach avoids threshold tuning, supports arbitrary sequence lengths via self-attention, and shows practical potential for end-to-end process mining applications. Future work includes extending to multi-view event attributes for richer anomaly modeling and repair.

Abstract

Event log records all events that occur during the execution of business processes, so detecting and correcting anomalies in event log can provide reliable guarantee for subsequent process analysis. The previous works mainly include next event prediction based methods and autoencoder-based methods. These methods cannot accurately and efficiently detect anomalies and correct anomalies at the same time, and they all rely on the set threshold to detect anomalies. To solve these problems, we propose a business process anomaly correction method based on Transformer autoencoder. By using self-attention mechanism and autoencoder structure, it can efficiently process event sequences of arbitrary length, and can directly output corrected business process instances, so that it can adapt to various scenarios. At the same time, the anomaly detection is transformed into a classification problem by means of selfsupervised learning, so that there is no need to set a specific threshold in anomaly detection. The experimental results on several real-life event logs show that the proposed method is superior to the previous methods in terms of anomaly detection accuracy and anomaly correction results while ensuring high running efficiency.

Figures (5)

• Figure 1: Overview of proposed method.
• Figure 2: Architecture of TransformerAE.
• Figure 3: Results of anomalous case classification task on BPIC2012.
• Figure 4: Results of anomalous correction task on BPIC2012.
• Figure 5: Comparison of anomaly correction tasks with baseline methods.