QUALITY ASSESSMENT OF PROCESS MODELS IN PROCESS MINING: THE CASE OF PETRI NETS

Abstract

Nowadays, information systems produce and consume enormous quantities of data. The data produced by Process Aware Information Systems is stored as event logs and can be converted into visual representations by using Process Mining algorithms. This paper follows a case study design with in-depth analysis of discovered Petri Nets’ quality. First, the theoretical dimensions of the research concerning the quality assessment of process models are presented, followed by a case study describing an electronic invoicing process. The process is analysed using discovery Process Mining algorithms having as output Petri Nets. The results of four Process Mining algorithms are analysed and qualitatively evaluated. Most process discovery algorithms consider only Fitness as qualitative criteria. Our study consider both Fitness and Precision metrics in the evaluation of discovered Petri Nets’ quality. The overall metric is also discussed. The findings of this study show similar percentages of Fitness and Precision for 3 out of 4 analysed models, the best overall percentage being 90%.