Discovering decision models from event logs

Abstract

Enterprise business process management is directly affected by how effectively it designs and coordinates decision making. To ensure optimal process executions, decision management should incorporate decision logic documentation and implementation. To achieve the separation of concerns principle, the OMG group proposes to use Decision Model and Notation (DMN) in combination with Business Process Model and Notation (BPMN). However, often in practice, decision logic is either explicitly encoded in process models through control flow structures, or it is implicitly contained in process execution logs. Our work proposes an approach of semi-automatic derivation of DMN decision models from process event logs with the help of decision tree classification. The approach is demonstrated by an example of a loan application in a bank.