Repairing Alignments of Process Models

Abstract

Process mining represents a collection of data driven techniques that support the analysis, understanding and improvement of business processes. A core branch of process mining is conformance checking, i.e., assessing to what extent a business process model conforms to observed business process execution data. Alignments are the de facto standard instrument to compute such conformance statistics. However, computing alignments is a combinatorial problem and hence extremely costly. At the same time, many process models share a similar structure and/or a great deal of behavior. For collections of such models, computing alignments from scratch is inefficient, since large parts of the alignments are likely to be the same. This paper presents a technique that exploits process model similarity and repairs existing alignments by updating those parts that do not fit a given process model. The technique effectively reduces the size of the combinatorial alignment problem, and hence decreases computation time significantly. Moreover, the potential loss of optimality is limited and stays within acceptable bounds.