Ver artículo

Process conformance is the domain within Process Mining that addresses the adequacy between a model and a log of a system. It has four different dimensions: fitness, precision, generalization and structure. This paper presents a metric to evaluate the precision dimension: the extra behavior a formal model tolerates when confronted to a log. Additionally, two important factors are presented together with the metric value: confidence (an estimation of the stability of the metric value for a future window), and a severity assessment to the imprecisions detected. Several techniques are described to accomplish this, including a log-guided traversal of the model, the optimization of binary linear programming models to estimate the confidence, and multi-factored methods to determine the severity. The approach is implemented within an open-source Process Mining platform, and experimental results certify both the significance of the approach and the usefulness of the new factors proposed.