This is the complementary repository for "Payment fraud identification by means of business process anomaly detection", the master thesis of Keyvan Amiri Elyasi submitted at Junior Professorship of Management Analytics, University of Mannheim (July 2022).
In this repository, synthetic datasets that are used for performance evaluation, as well as sample GAE-GAT, BINet, and GAE-ECC models trained on these datasets can be found. More importantly, Jupyter Notebooks that are used for data preprocessing, data transformation, training, hyperparameter optimization, and performance evaluation are included in the repository. For the sake of brevity, Jupyter Notebooks which start by ”GAE-GAT-01” to ”GAE-GAT-06” only include computations for some of the datasets. For other dataset the same procedure is applied. Other Jupyter Notebooks which focus on performance analysis include results that are discussed in the master thesis report.