Identifying Money Laundering Accounts

Abstract

Money laundering is often associated with criminal activities. Anti-money laundering is thus regarded as an important task in many countries. However, as it is common that money launderers divide the dirty money into multiple parts and make sequences of banking transfers or commercial transactions, manually detecting activities of money laundering is challenging. To ease the task, this work establishes a two-phase intelligent method based on machine learning and data analysis techniques for identifying suspicious money laundering accounts from the transaction data. The first phase emphasizes on identifying every suspicious money laundering account while the second phase further retrieves highly suspicious ones so that both the recall and precision for the identification of money laundering accounts can be somewhat taken care of. Evaluated on the data given by Bank SinoPac, the established intelligent method achieves a recall rate of 26.3%, which is three times the recall rate (8.6%) of the Money Laundering Control Act in Taiwan, in the first phase, and later the precision rate can be increased up to 87.04% in the second phase.