Enhancement in Process Mining Model by Repairing Noisy Behavior in Event Log

Abstract

Companies and organizations aim to improve the performance of their business processes to stay competitive. Recently, researchers have shown significant interest in process mining, particularly its ability to extract accurate information from process-related data. Process enhancement is a crucial aspect of process mining, involving the extraction of information from the actual process event log to extend or improve existing processes. Enhancement can be classified into two types: extension and repair. This paper specifically focuses on the repair type of enhancement. Information systems commonly encounter logging errors or exhibit special behaviors that introduce noise into the event log. In this research, we investigate the process mining model in the presence of noise in the event log. We propose a method for repairing event logs by decomposing them into sub-logs and eliminating the noisy behavior within these sub-logs using covering probability. The repaired sub-logs are then integrated into the original event log at the appropriate location. Additionally, we propose a probabilistic method that considers the frequency of occurrence for activities in a given situation. This method allows for the removal of noisy and abnormal behavior from the event log, providing an overall perspective on the process. To validate our approach, we generate artificial event logs with the presence of noisy behavior using the ProM framework. By using the RapidMiner-based ProM Extension, we generate a test set to illustrate how various types of noisy behavior in an event log can be identified and repaired. Our findings clearly demonstrate that repairing the event log improves the performance of a process mining model.

Keywords

• Covering probability
• enhancement
• outliers
• process mining
• repaired event log