Data engineering for fraud detection

Financial institutions increasingly rely upon data-driven methods for developing fraud detection systems, which are able to automatically detect and block fraudulent transactions. From a machine learning perspective, the task of detecting suspicious transactions is binary classification problem therefore many techniques can be applied. Interpretability however utmost importance management have confidence in model designing prevention strategies. Moreover, models that enable experts understand underlying reasons why case flagged as will greatly facilitate their job investigating Therefore, we propose several data engineering improve performance an analytical while retaining interpretability property. Our process decomposed into feature instance steps. We illustrate improvement these steps popular on real payment set.