The code for preprocessing event logs is located within the Whole_Graph and Subgraph modules.
- Whole_Graph: This module is responsible for processing the entire raw event log data and constructing a multi-layer heterogeneous graph based on the complete event logs.
- Subgraph: This module is used to process the multi-layer heterogeneous graph based on the complete event logs, from which it extracts isomorphic and heterogeneous subgraphs.
To train, validate, and test the model, use the following command (example for event log BPI Challenge 2013 closed_problems):
All preprocessing of event logs and training of models are consolidated into a single script, main.py
, for streamlined execution. This script allows you to choose the dataset you wish to test against.
You can specify the dataset within the main.py
script. As an example, the bpic2017_o
event log dataset can be selected for testing.
To run the preprocessing and model training on your chosen dataset, execute the following command in your terminal:
python main.py
This command will trigger the processes as defined in the main.py script for bpic2017_o
event log.
To load a saved model and test it on the test set, run (example for event log bpic2017_o
):
python metrics.py
pytorch: Used for deep learning operations. python: The programming language used for the project.
The event logs for predictive business process monitoring can be found at 4TU Research Data.
This repository also includes implementations of various baseline methods for predictive business process monitoring, including MiDA, ProcessTransformer, MiTFM, PREMIERE, gcn-procesprediction, BIGDGCNN, and Multi-BIGDGCNN.
- Reference: Vincenzo Pasquadibisceglie, Annalisa Appice, Giovanna Castellano, and Donato Malerba. A multi-view deep learning approach for predictive business process monitoring. IEEE Transactions on Services Computing, 15(4):2382–2395, 2022.
- Reference: Zaharah A Bukhsh, Aaqib Saeed, and Remco M Dijkman. Processtransformer: Predictive business process monitoring with transformer network. arXiv preprint arXiv:2104.00721, 2021.
- Reference: Jiaxing Wang, Chengliang Lu, Bin Cao, and Jing Fan. MiTFM: A multi-view information fusion method based on transformer for next activity prediction of business processes. In Proceedings of the 14th Asia-Pacific Symposium on Internetware, pages 281–291, 2023.
- Reference: V. Pasquadibisceglie, A. Appice, G. Castellano, and D. Malerba. Predictive process mining meets computer vision. In Business Process Management Forum: BPM Forum 2020, Seville, Spain, September 13–18, 2020, Proceedings 18, pages 176–192, 2020.
- Reference: Ishwar Venugopal, Jessica Töllich, Michael Fairbank, and Ansgar Scherp. A comparison of deep-learning methods for analysing and predicting business processes. In 2021 International Joint Conference on Neural Networks (IJCNN), pages 1–8. IEEE, 2021.
- Reference: Andrea Chiorrini, Claudia Diamantini, Alex Mircoli, and Domenico Potena. Exploiting instance graphs and graph neural networks for next activity prediction. In International conference on process mining, pages 115–126, 2021.
- Reference: Andrea Chiorrini, Claudia Diamantini, Laura Genga, and Domenico Potena. Multi-perspective enriched instance graphs for next activity prediction through graph neural network. Journal of Intelligent Information Systems, 61(1):5–25, 2023.