Title:Detection of statistically significant differences between process variants through declarative rules

Abstract:Services and products are often offered via the execution of processes that vary according to the context, requirements, or customisation needs. The analysis of such process variants can highlight differences in the service outcome or quality, leading to process adjustments and improvement. Research in the area of process mining has provided several methods for process variants analysis. However, very few of those account for a statistical significance analysis of their output. Moreover, those techniques detect differences at the level of process traces, single activities, or performance. In this paper, we aim at describing the distinctive behavioural characteristics between variants expressed in the form of declarative process rules. The contribution to the research area is two-pronged: the use of declarative rules for the explanation of the process variants and the statistical significance analysis of the outcome. We assess the proposed method by comparing its results to the most recent process variants analysis methods. Our results demonstrate not only that declarative rules reveal differences at an unprecedented level of expressiveness, but also that our method outperforms the state of the art in terms of execution time.