Detecting financial frauds: the role of anomaly detection in AML pipeline

In the landscape of sophisticated financial crimes, countering money laundering and terrorist financing has become a critical issue. Transaction monitoring, a pivotal component in this endeavor, faces its share of hurdles. The complexities inherent in modern banking, digital transactions, and the expansion of global systems provide a fertile ground for criminals to exploit vulnerabilities, thereby amplifying the intricacy of detection. This thesis takes on the task of navigating these intricate challenges by delving into the pivotal role that transaction monitoring assumes. Central to this exploration is the introduction of customer segmentation and threshold calibration, concepts poised to enhance the efficacy of transaction monitoring systems. These mechanisms are examined for their potential to bolster the overall robustness of the monitoring process. This study presents and utilizes the method of the Isolation Forest, an advanced machine learning technique for outlier detection. This method offers a promising avenue toward addressing the aforementioned threats, adding to the arsenal of potential solutions in this ongoing battle.