On the Representational Bias in Process Mining

Abstract

Process mining serves a bridge between data mining and business process modeling. The goal is to extract process related knowledge from event data stored in information systems. One of the most challenging process mining tasks is process discovery, i.e., the automatic construction of process models  from raw event logs. Today there are dozens of process discovery techniques generating process models using different notations (Petri nets, EPCs, BPMN, heuristic nets, etc.). This paper focuses on the representational bias used by these techniques. We will show that the choice of target model is very important for the discovery process itself. The representational bias should not be driven by the desired graphical representation but by the characteristics of the underlying processes and process discovery techniques. Therefore, we analyze the role of the representational bias in process mining.