A Visual Approach Based on Petri Net to Compare Business Processes

 

MSc Thesis Project

Title: A Visual Approach Based on Petri Net to Compare Business Processes

Author: Yan Wo

Daily Supervisor: Dr. Sebastiaan J. van Zelst

1st examiner: Prof.Dr.ir Wil M.P. van der Aalst

2nd examiner: Prof.Dr. Peter Rossmanith

Summary

In various application domains, comparing the behavior of different processes is of interest, e.g., to analyze the shopping habits of consumers from different countries, or to compare the execution of two computer programs to improve the general performance. This thesis addresses the problem of comparing two processes represented by event logs. Some recent approaches that aim to solve this problem either only give a binary answer, or are too complex to implement, or fail to display the discovered differences in an intuitive and readable way. Furthermore, how to account for concurrent behavior in the processes comparison is a problem as well. In this thesis, we aim to detect and visualize overflow differences between processes, i.e., the occurrence frequency of the activities and the causal relations between activities (i.e., which activities are typically observed to follow given activities). We propose a new method to construct a Petri net based on two given event logs and highlight the differences in the Petri net with the help of diverging colors and labels. By performing experiments with real-life event data, we validate and evaluate our approach. The experiment results show how our approach detects and visualizes the over-flow differences between two processes recorded in event logs and how our approach captures concurrent behaviors that are missed by previous approaches.