Comparative analysis of machine learning algorithms for money laundering detection

This study explored the effectiveness of anomaly detection techniques in identifying fraudulent financial transactions, with a particular focus on money laundering activities. The research addressed the growing challenges of financial fraud, which significantly impacts economies and financial institutions by leading to substantial monetary losses and undermining trust in financial systems. This research examined contemporary machine learning (ML) algorithms, including XGBoost, K-Nearest Neighbors, Random Forest, Isolation Forest, and Support Vector Machines, to analyze transaction data for anomalies indicative of fraudulent behavior. This study’s approach involves data collecting, system design, implementation, data analysis, and experimental setup. This research aimed to identify robust algorithms for detecting financial fraud. In the results notably, XGBoost shows an output of 1.0, 1.0, 1.0, 1.0, and 0.94 for accuracy, precision, recall, F1 score and AUC respectively to outperform other ML algorithms experimented in money laundering detection. The findings underscore the potential of ML algorithms capability to combat money laundering and indeed enhance anti-money efforts and offer financial institutions a powerful, resource-efficient method for fraud detection in large-scale transaction environments.