Fair Fraud Detection

about.

Fair Fraud Detection Dashboard for Adyen

As part of a team of Information Systems Master’s students, we develop a data-driven dashboard for Adyen to support fair and customizable fraud detection. Using stakeholder insights and iterative experimentation, we built an XGBoost-based model that lets merchants adjust the sensitivity of fraud flagging. This gives them control over balancing fraud prevention with customer satisfaction.

Fair Fraud Detection Dashboard for Adyen

To ensure fairness and flexibility in fraud detection, we explored both knowledge graphs and XGBoost, ultimately selecting XGBoost for its effectiveness. The model was trained on a balanced dataset using oversampling to address class imbalance. Multiple versions of the model were created to let merchants choose their preferred trade-off between false positives and false negatives—allowing them to prioritize either accuracy or customer fairness.

The interactive dashboard we built allows merchants to adjust the fraud detection trade-off and view recent transactions alongside key fraud metrics. We recommended continued use of XGBoost for its accuracy, fairness, and customizability. Future improvements—such as external user validation, RDF/SPARQL integration, and enhanced visualizations—could further increase transparency and strategic value for merchants.