An empirical evaluation of process mining algorithms based on structural and behavioral similarities

While many process mining algorithms have been proposed recently, there exists no widely-accepted benchmark to evaluate these process mining algorithms. As a result, it can be difficult to compare different process mining algorithms especially over different application domains. This paper presents our attempt in building such a benchmark by empirically evaluating process mining algorithms using reference models, in which the quality of a discovered model is measured by the behavioral and structural similarities with its reference model. In addition to artificial reference models extracted from academic papers and SAP suites, real-life processes from a major boiler manufacturer in China are added into the benchmark.