Fork of Preserving Complex Object-Centric Graph Structures to Improve Machine Learning Tasks in Process Mining

Jan Niklas Adams, Gyunam Park, Wil van der Aalst forked by Paul Taurand

The experiments are based on the Python library ocpa

---

Instructions

Use venv to create the environment.

python -m venv /path/to/new/virtual/environment.

Activate the environment

/path/to/new/virtual/environment/.venv/Script/activate

Install all the dependencies from requirements.txt

pip install -r requirements.txt

Go into the repository directory and unzip BPI2017-Final.zip before run

python main_BN.py

This will run the code for prediction with bayesian network. The output is the file results_BN\metrics_BN.csv. To vizualize the curve on the graph of the initial projet, you need to run the notebook Visualization_BN.ipynb the documention is in.

---

Notes

• The original code is used to work with conda
• All files followed by _BN are files modified by the Fork.
• In main.py, you'll find the if statement for Bayesian network prediction. However, the code fails to make all predictions.
• A change has been made in the gnn_utils.py file between lines 101 and 240. A library problem was preventing main_BN.py, which was the aim of this project, from working properly.