Discovering duplicate tasks in transition systems for the simplification of process models

Abstract

This work presents a set of methods to improve the understandability of process models. Traditionally, simplification methods trade off quality metrics, such as fitness or precision. Conversely, the methods proposed in this paper produce simplified models while preserving or even increasing fidelity metrics. The first problem addressed in the paper is the discovery of duplicate tasks. A new method is proposed that avoids overfitting by working on the transition system generated by the log. The method is able to discover duplicate tasks even in the presence of concurrency and choice. The second problem is the structural simplification of the model by identifying optional and repetitive tasks. The tasks are substituted by annotated events that allow the removal of silent tasks and reduce the complexity of the model. An important feature of the methods proposed in this paper is that they are independent from the actual miner used for process discovery.