Anti-Money Laundering Risk Identification of Financial Institutions based on Aspect-Level Graph Neural Networks

Abstract

The contemporary financial industry is a highly information-based industry. The digital system can establish a complete information system around various attributes and behaviors of bank accounts. In the core business system, most of this information is constantly changing and recorded in real time. Therefore, we can achieve the goal of monitoring the money laundering risk of the account by analyzing the relevant element data and specific characteristics of the account. The risk assessment and customer classification indicator system for accounts is composed of four basic elements: customer characteristics, location, business development and industry conditions. Account money laundering risk indicators are composed of various basic elements and their risk sub-items. We propose an aspect-based (aspect-level) graph convolutional neural network, starting from different perspectives, to quantify the risk of money laundering in financial institutions.