On the Performance Analysis of the Adversarial System Variant Approximation Method to Quantify Process Model Generalization

Process mining algorithms discover a process model from an event log. The resulting is supposed to describe all possible sequences of the underlying system. Generalization quality dimension interest. A generalization metric should quantify extent which represents observed contained in log and unobserved Most available metrics literature cannot properly model. recently published method [1] called Adversarial System Variant Approximation leverages Generative Networks approximate sequence distribution system While this demonstrated performance gains over existing methods measuring models, its experimental evaluations have been performed under ideal conditions. This paper experimentally investigates non-ideal conditions such as biased limited logs. Moreover, experiments are investigate originally proposed sampling hyperparameter value on measure generalization. results confirm need raise awareness about working method. outcomes also serve initiate future research directions. Theis, Julian, Houshang Darabi. "Adversarial Quantify Model Generalization." IEEE Access 8 (2020): 194410-194427.