IBM反洗钱数据集

本数据集模拟虚拟全球金融生态中的个体、企业与银行间交过程互，涵盖合法交易及完整的洗钱周期(注入、混合、整合)，旨在为反洗钱模型训练、异常检测与图神经网络研究提供带标签的合成基准数据，数据基于IBM参数化交易模拟器生成，时间范围为2022年8-11月(日级精度，部分至小时级)，通过实体建模、行为分布驱动的交互模拟、资金流追踪与模式注入实现加工，提供高洗钱比例(HI)和低洗钱比例(LI)两类多种规模子集，支持跨机构全视图分析与可复现实验。This dataset simulates the interaction processes among individuals, enterprises and banks within a virtual global financial ecosystem, covering legitimate transactions and the complete money laundering cycle (placement, layering, integration). It is designed to provide labeled synthetic benchmark data for anti-money laundering (AML) model training, anomaly detection and graph neural network research. The dataset is generated based on the IBM Parametric Transaction Simulator, with a time span from August to November 2022 (daily granularity, with partial data at hourly granularity). It is processed through entity modeling, behavior distribution-driven interaction simulation, fund flow tracking and pattern injection, and offers multiple scale subsets under two categories: high money laundering ratio (HI) and low money laundering ratio (LI). It supports cross-institution full-view analysis and reproducible experiments.

搜集汇总

背景与挑战

背景概述

IBM反洗钱数据集是一个合成基准数据集，模拟了全球金融生态中的交易过程，涵盖合法交易和完整的洗钱周期（注入、混合、整合），旨在为反洗钱模型训练、异常检测和图神经网络研究提供带标签的数据。该数据集基于IBM参数化交易模拟器生成，时间范围为2022年8-11月，提供高洗钱比例和低洗钱比例两类多种规模子集，支持跨机构全视图分析和可复现实验。