# An End-to-End Approach for Online Decision Mining and Decision Drift Analysis in Process-Aware Information Systems: Extended Version

arXiv: https://arxiv.org/abs/2303.03961 · HTML: https://mlsanity.com/paper/2303.03961/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Builds on 3 of its 27 references that are in the map.
- First posted on arXiv 7 Mar 2023.

## Builds on (most cited first, 3 of 3)

- [Event Stream-Based Process Discovery using Abstract Representations](https://mlsanity.com/paper/1704.08101/) (2017)
- [Explainable Predictive Decision Mining for Operational Support](https://mlsanity.com/paper/2210.16786/) (2022)
- [Detecting and Understanding Branching Frequency Changes in Process Models](https://mlsanity.com/paper/2103.10742/) (2021)

## Built on by (most cited first, 2 of 2)

- [Tailoring Machine Learning for Process Mining](https://mlsanity.com/paper/2306.10341/) (2023)
- [Online Discovery of Simulation Models for Evolving Business Processes (Extended Version)](https://mlsanity.com/paper/2506.10049/) (2025)

## Abstract (arXiv)

Decision mining enables the discovery of decision rules from event logs or streams, and constitutes an important part of in-depth analysis and optimisation of business processes. So far, decision mining has been merely applied in an ex-post way resulting in a snapshot of decision rules for the given chunk of log data. Online decision mining, by contrast, enables continuous monitoring of decision rule evolution and decision drift. Hence this paper presents an end-to-end approach for the discovery as well as monitoring of decision points and the corresponding decision rules during runtime, bridging the gap between online control flow discovery and decision mining. The approach provides automatic decision support for process-aware information systems with efficient decision drift discovery and monitoring. For monitoring, not only the performance, in terms of accuracy, of decision rules is taken into account, but also the occurrence of data elements and changes in branching frequency. The paper provides two algorithms, which are evaluated on four synthetic and one real-life data set, showing feasibility and applicability of the approach. Overall, the approach fosters the understanding of decisions in business processes and hence contributes to an improved human-process interaction.