Estimating Process Conformance by Trace Sampling and Result Approximation

Abstract

The increasing volume of event data that is recorded by information systems during the execution of business processes creates manifold opportunities for process analytics. Specifically, conformance checking compares the behaviour as recorded by an information system to a model of desired behaviour. Unfortunately, state-of-the-art conformance checking algorithms scale exponentially in the size of both the event data and the model used as input. At the same time, event data used for analysis typically relates only to a certain interval of process execution, not the entire history. Given this inherent data incompleteness, we argue that an understanding of the overall conformance of process execution may be obtained by considering only a small fraction of a log. In this paper, we therefore present a statistical approach to ground conformance checking in trace sampling and conformance approximation. This approach reduces the runtime significantly, while still providing guarantees on the accuracy of the estimated conformance result. Comprehensive experiments with real-world and synthetic datasets illustrate that our approach speeds up state-of-the-art conformance checking algorithms by up to three orders of magnitude, while largely maintaining the analysis accuracy.