Detection of Suspicious Financial Transactions in the Banking Domain to Counter Money Laundering

Abstract

Money laundering is a serious offense in many countries including Pakistan. In Pak istan, money laundering is a major problem that is frequently linked to corruption, drug trafficking, and support for terrorism. The State Bank of Pakistan and the Financial Monitoring Unit are present, but the country’s regulatory and legislative framework to combat money laundering is weak. The efficiency of these institutions is hampered by inadequate budget, a lack of political backing, and internal corruption. A nation’s capital, cash flow, and financial assets are badly impacted by financial fraud, which includes banking frauds like credit card fraud and internet banking fraud. It also funds illegal operations. Banks and other financial institutions develop anti-money laundering (AML) policies and processes to stop money laundering. This study aims to simulate real-world banking transactions using a synthetic data-set in order to detect illicit ac tivities. By employing data-driven machine learning and deep learning techniques, the study classifies fraudulent transactions and presents the results. Multiple models are trained, tested, and compared to determine their performance. Additionally, the study explores the factors contributing to fraudulent transactions by analyzing LIME values. Based on the findings, the study recommends the best-performing models for identifying suspicious transactions.