Deep Q-network-based adaptive alert threshold selection policy for payment fraud systems in retail banking

Abstract

Machine learning models have widely been used in fraud detection systems. Most of the research and development efforts have been concentrated on improving the performance of the fraud scoring models. Yet, the downstream fraud alert systems still have limited to no model adoption and rely on manual steps. Alert systems are pervasively used across all payment channels in retail banking and play an important role in the overall fraud detection process. Current fraud detection systems end up with large numbers of dropped alerts due to their inability to account for the alert processing capacity. Ideally, alert threshold selection enables the system to maximize the fraud detection while balancing the upstream fraud scores and the available bandwidth of the alert processing teams. However, in practice, fixed thresholds that are used for their simplicity do not have this ability. In this paper, we propose an enhanced threshold selection policy for fraud alert systems. The proposed approach formulates the threshold selection as a sequential decision making problem and uses Deep Q-Network based reinforcement learning. Experimental results show that this adaptive approach outperforms the current static solutions by reducing the fraud losses as well as improving the operational efficiency of the alert system.

Designing an Intelligent Scoring System for Crediting Manufacturers and Importers of Goods in Industry 4.0

“…In this setting of credit cards, blockchain technology provides enhanced security, transparency, and trust in financial transactions. It reduces the risk of fraud, minimizes the need for intermediaries, and ensures that transaction histories are reliable and immutable [39][40][41][42]. Manufacturers, importers, and financial institutions can benefit significantly from these blockchain advantages in credit card transactions [43][44][45][46].…”

DeepTOP: Deep Threshold-Optimal Policy for MDPs and RMABs

“…[13] proposed an RL-based threshold policy for semi-MDPs in controlling micro-climate for buildings with simulations proving efficacy on a single-zone building. [29] used the Deep Q-network RL algorithm for selecting alert thresholds in anti-fraud systems with simulations showing performance improvements over static threshold policies. [26] described the SALMUT RL algorithm for exploiting the ordered multi-threshold structure of the optimal policy with SALMUT implementations in [16] for computing node's overload protection.…”

Graph Computing for Financial Crime and Fraud Detection: Trends, Challenges and Outlook

“…This changes the temporal characteristics of the processing task significantly. However, for most channels, due to the large number of alerts generated, manual alert processing systems may face bandwidth limitations [80]. Prioritizing the accounts based on the risk assessment through graph analysis provides a solution towards lower financial losses and improved response times.…”