A Suspicious Financial Transaction Detection Model Using Auto encoder and Risk-Based Approach

This research addresses the identification of suspicious financial transactions within the increasingly complex and opaque landscape shaped by modern electronic financial technologies. It introduces a model designed to dynamically capture trends in diverse fund and transaction flows, with the capability to autonomously learn intricate transaction patterns. Central to this approach is the development of an internal control system based on a risk-focused methodology, which leverages auto encoders to strengthen anti-money laundering (AML) practices. This strategy demonstrates superior performance compared to conventional AML techniques. Furthermore, the model aids in identifying potential indicators of suspicious activity and supports the continuous refinement of alert systems within AML frameworks. This enables more efficient case analysis and reduces the need for exhaustive transaction reviews. Overall, the proposed model significantly enhances both the effectiveness and efficiency of AML operations, offering valuable insights for improving financial institutions' systems for evaluating and managing the risks associated with suspicious transactions.