DyLoPro: Profiling the Dynamics of Event Logs - Case Studies

This repository contains numerous annotated notebooks with an extensive and comprehensive analysis of the dynamics over time for multiple public event logs commonly used in Process Mining (PM) literature. As part of an ongoing effort, the repository will continue to grow with additional case studies of additional public event logs over time. These case studies are conducted leveraging the powerful capabilities of the newly released DyLoPro Python library.

1. DyLoPro Python Library

The DyLoPro Python library is a tool that allows PM practitioners to efficiently and comprehensively explore the dynamics in event logs over time, prior to applying PM techniques. These comprehensive exploration capabilities are provided by extensive set of plotting functionalities, enabling PM researchers and practitioners to visualize the dynamics over time from different process perspectives.

For more information about the DyLoPro library, please refer to the project description, which can be found on both the DyLoPro Github repository and on DyLoPro's PyPI page. The project description provides numerous useful resources, including, inter alia, links to the documentation and comprehensive instructions on how to get started.

In addition to the package documentation and instructions, this repository's detailed notebooks comprising the case studies conducted on a number of commonly used real-life event logs might also significantly improve your understanding on how to use and access DyLoPro's variety of plotting methods.

2. Extensive case studies public event logs

Most Process Mining techniques assume that the underlying process and hence the data generating function of the associated event log is in a steady-state. However, in reality, this is rarily the case.

"Applying PM on event logs in which this stationarity assumption does not hold, i.e. in which one or more drifts occur in the underlying process, can induce a significant yet oftentimes unnoticed bias in the results, leading to incorrect insights."

New PM techniques proposed in the literature are frequently evaluated and compared using commonly employed public event logs. However, potential sources of bias stemming from underlying event log drifts are almost always overlooked. Therefore, case studies are conducted for a number of these commonly used public event logs. They demonstrate both the usefulness and effectiveness of leveraging the DyLoPro framework and library for comprehensively analyzing the dynamics over time from multiple perspectives. Each case study is documented in an annotated notebook that uncovers the most interesting trends / patterns and changes over time for the associated event logs.

These case studies significantly enhances the transparency of several public event logs widely employed in the Process Mining literature, thereby contributing to the advances in the field of PM in several ways:

1. Enabling Process Mining researchers to interpret the results of their proposed techniques when applied on these logs in a more informed manner.
2. The differences in results between techniques can be identified with a better understanding of potential sources of bias.
3. Enabling researchers to make more informed decisions on how to preprocess the event logs and take appropriate actions to address any patterns or drifts that may induce bias in the results.
4. Assisting researchers in potentially even determining how to subset the data to avoid bias.

The overall advantages of the increased transparency provided by these case studies, and hence provided by the DyLoPro package in general, demonstrate its potential to significantly improve the quality and accuracy of Process Mining research and contribute to advancing the field.

"The impact of drifts in different perspectives on the results of the process mining techniques varies depending on the technique used."

The primary aim of DyLoPro is not to provide a conclusive answer on the root causes and consequences of certain drifts, but rather to enable researchers and practitioners to efficiently explore a wide range of dynamics present in the often very complex data structures that event logs are, and accordingly efficiently identify any changes, trends or patterns of interest for subsequent analysis. Researchers might or might not decide to further analyze certain identifed drifts or patterns, depending on the Process Mining task at hand.

The associated annotated notebooks are:

1. BPIC_19.ipynb: Comprehensive case study on the dynamics over time in the BPIC19 event log. The data can be found here.
2. BPIC_17.ipynb: Comprehensive case study on the dynamics over time in the BPIC17 event log. The data can be found here.
3. BPIC_15.ipynb: Comprehensive case study on the dynamics over time in the BPIC15 event log. The data can be found here. COMING SOON...
4. BPIC_12.ipynb: Comprehensive case study on the dynamics over time in the BPIC12 event log. The data can be found here. COMING SOON...
5. BPIC_20: BPIC_20 event log is a collection of 5 event logs pertaining to the travel administration at a university. Each event log covers the cases of a different process. An overview of the data can be found here.
   1. BPIC_20_DomDecl.ipynb: Comprehensive case study on the dynamics over time in the Domestic Declaration event log. The data can be found here. COMING SOON...
   2. BPIC_20_IntlDecl.ipynb: Comprehensive case study on the dynamics over time in the International Declaration event log. The data can be found here. COMING SOON...
   3. BPIC_20_PpTravel.ipynb: Comprehensive case study on the dynamics over time in the Prepaid Travel Costs event log. The data can be found here. COMING SOON...
   4. BPIC_20_TravelPermit.ipynb: Comprehensive case study on the dynamics over time in the Travel Permit Data event log. The data can be found here. COMING SOON...
   5. BPIC_20_RequestPay.ipynb: Comprehensive case study on the dynamics over time in the Request For Payment event log. The data can be found here. COMING SOON...
6. Road_Traffic.ipynb: Comprehensive case study on the dynamics over time in the Road Traffic Fines Management event log. The data can be found here. COMING SOON...
7. Hospital_Billing.ipynb: Comprehensive case study on the dynamics over time in the Hospital Billing event log. The data can be found here. COMING SOON...