A rule-based framework for interpretable predictions of business process outcomes using event logs

Abstract

The development of models for process outcome prediction using event logs has evolved with a clear focus on performance improvement. In this thesis we take a different perspective, focusing on obtaining interpretable predictive models for outcome prediction. In particular, we propose a method based on association rule mining, which results in inherently interpretable classification models. While association rule mining has been used with event logs for process model approximation and anomaly detection in the past, its application to outcome-based predictive model is novel. The proposed method defines how to pre-process logs, obtain the rules, prune the rules to a limited number that can be handled by human decision makers, and use the rules to predict process outcomes. The experimental results on real world event logs show that in most cases the performance of the proposed method is aligned with the one of traditional approaches, with only a slight decrease in some cases. We argue that such a decrease of performance is an acceptable trade-off in return for a predictive model that is interpretable by design.