Adaptive Transaction Sequence Neural Network for Enhanced Money Laundering Detection

This paper introduces a novel approach to money laundering detection by leveraging the adaptive characteristics of financial transactions. The ATSNN model addresses the limitations of traditional static analysis methods by incorporating adaptive behavioral and structural features extracted from transaction data. It captures the evolving patterns and behaviors indicative of money laundering. A feature enhancement module is utilized to uncover the latent interplay between behavioral patterns and structural evolution, enriching the feature set with an enhancement vector derived from the learned adaptive features. The paper details the methodology and architecture of the ATSNN, demonstrating its capability to effectively integrate and analyze adaptive transaction features. Extensive experiments on real-world datasets substantiate the model's efficacy in discerning money laundering accounts from benign ones. The results emphasize the importance of adaptive feature extraction in improving the detection of illicit financial activities, representing a significant advancement in anti-money laundering technologies.