Triggering Proactive Business Process Adaptations via Online Reinforcement Learning

by Andreas Metzger, Tristan Kley, and Alexander Palm

For the sake of transparency, reproducibility, and replicability, below we provide the links to the artefacts created as part of the work on our BPM 2020 submission in order to facilitate reproducibility and replicability.

Code

• Code to train and evaluate the RNN-LSTM ensembles: https://github.com/Chemsorly/BusinessProcessOutcomePrediction

• Code to generate predictions and reliability estimates, linked to the RL algorithm: https://github.com/Trikley/proactive_BPM_adaptation

Data Sets

We used the real-world data sets Traffic, BPIC2012 and BPIC2017 for experimental evaluation

• Original source: https://data.4tu.nl/repository/

• Transformed into the input for RNN-LSTM models: https://uni-duisburg-essen.sciebo.de/s/zpX4kRXkgcX3Ez4

• Prediction results for all RNN-LSTM models of the ensemble: https://uni-duisburg-essen.sciebo.de/s/iFy6y0BsAWTgWLV

Experimental Results

• Available from: https://uni-duisburg-essen.sciebo.de/s/6JyqueaMBQmHMXt