Complex Control-Flow Constructs Detection from Process Related Data

Abstract

Process mining is new techniques whereby knowledge from event log stored in today’s information systems are extracted to automatically construct business process models to have a full understanding of the real behaviour of processes, identify bottlenecks, and then improve them. Many process discovery algorithms have been proposed today. However, there are complex control-flow constructs that current discovery techniques cannot correctly discover in models based on event logs. These constructs are short loops, invisible tasks, and nonfree choice constructs. There is currently no algorithm that can handle all of these structures in a restricted time. In this paper, we propose a framework that detects from event logs the complex control-flow constructs that exist. By identifying the existing constructs from a given event log, one can identify the process model notation or the process discovery algorithm appropriate for the given event log. The framework has been implemented in ProM and the results show that constructs are identified correctly.