Explainable Generative Models in FinCrime
Ankur Mahida
Abstract
Explainable generative models (EGMs) are a strong tool in the crime domain that is transparent with the combination of generative models that learn complex data distributions with interpretability methods.This review covers how EGMs have been used to fight financial crimes, including money laundering, fraud, terrorist financing, etc. EGMs can provide a solution by opening the model's inner workings through explain ability methods to meet these challenges.The review examines the key differentiators between EGMs.It explains their uses (AML, fraud detection, and FinCrime in general) by showing the possible ways they change the status quo in the financial sector and regulators.EGMs incorporate saliency maps, attention mechanisms, and counterfactual explanations to give human-terminating revelations, which, in return, builds trust and empowers effective decision-making in FinCrime utilization.