Open AccessAutomatic discovery of failures in business processes using Process Mining techniques
Author(s) -
Guillermo Calderón-Ruiz,
Marcos Sepúlveda
Publication year - 2013
Language(s) - English
Resource type - Conference proceedings
DOI - 10.5753/sbsi.2013.5710
Subject(s) - computer science , process mining , business process , business process discovery , event (particle physics) , process (computing) , data mining , control flow , business process management , perspective (graphical) , control (management) , business process modeling , data science , artificial intelligence , work in process , engineering , operations management , programming language , physics , quantum mechanics , operating system
One of the most common and costly problems that organizations are facing is to find the causes of failures in business processes. Failures are often due to missing or unnecessary execution of some process activities; or with how some activities are performed. Currently, there is no automatic technique that helps finding these causes. We propose a novel technique to identify potential causes of failures in business process by extending available Process Mining techniques. Initially, the original event log is filtered in two logs, the former with successful cases and the latter with failed cases. Then, behavioral patterns are extracted from both event logs using the Performance Sequence Diagram Analysis algorithm. Finally, both sets of patterns are compared considering control flow and time perspectives. The differences found represent potential causes of failures in business processes. We tested this technique using several synthetic event logs. Results show the technique is able to successfully find missing or unnecessary activities, and failed behavioral patterns that differ from successful patterns either in the control flow or in the time perspective.