Improving Anti Money Laundering through Active learning Methods

Fighting against financial crime is a battle that humanity is losing as a whole. Money Laundering is the critical component of any criminal activity that needs to be profitable. Therefore, fighting money laundering is fighting against all crimes including terrorism, human trafficking, drugs and wildlife trafficking. The United Nations (UN) estimates 2 trillion USD is laundered globally, and only less than 0.2% of that can be recovered by the authorities. The main responsibility lies in the financial institutions, where the majority of the laundering activity occurs. However, these institutions use rule-based systems that generate huge amount false-positive alerts and fail to catch criminals. Machine learning is a promising tool that has shown good results for complex tasks such as this one. In this project we propose to tackle the problem of money laundering detection through active learning strategies, where a human expert provides labels for cases carefully selected by the machine learning algorithm. We will identify the most well suited active learning strategies for the problem using multiple types of data including individual customer data and sub-graphs. A new special-purpose active learning solution will be developed to achieve better performance for this problem

Faculty Supervisor:

Paula Branco

Student:

Partner:

H3M Analytics Inc.

Discipline:

Computer science

Sector:

Professional, scientific and technical services

University:

University of Ottawa

Program: