Repairing Alignments of Process Models

Process mining represents a collection ofdata driven techniques that support the analysis, un-derstanding and improvement of business processes. Acore branch of process mining is conformance checking,i.e., assessing to what extent a business process modelconforms to observed business process execution data.Alignments are the de facto standard instrument to com-pute such conformance statistics. However, computingalignments is a combinatorial problem and hence ex-tremely costly. At the same time, many process modelsshare a similar structure and/or a great deal of behavior.For collections of such models, computing alignmentsfrom scratch is inefficient, since large parts of the align-ments are likely to be the same. This paper presentsa technique that exploits process model similarity andrepairs existing alignments by updating those parts thatdo not fit a given process model. The technique effect-ively reduces the size of the combinatorial alignmentproblem, and hence decreases computation time signi-ficantly. Moreover, the potential loss of optimality islimited and stays within acceptable bounds.