Machine Learning for Money Laundering Analysis

This study investigates the application of machine learning (ML) for proactive threat hunting in financial transaction systems, focusing on detecting money laundering activities—a critical challenge in business cybersecurity. Utilizing the Elliptic Bitcoin dataset, we apply six supervised ML models: Random Forest, Support Vector Machine, K-Nearest Neighbors, Stochastic Gradient Descent, Decision Tree, and Logistic Regression. We address class imbalance using the Synthetic Minority Oversampling Technique (SMOTE) and conduct exploratory data analysis to identify key transactional features. Evaluated using cybersecurity-relevant metrics (accuracy, precision, recall, F1-score), the Random Forest classifier outperforms others, achieving 98.87% accuracy on the holdout set. This work contributes an empirical benchmarking study demonstrating that, on the Elliptic Bitcoin dataset, Random Forest achieves the highest macro-F1 (0.965) and ROC-AUC (0.996) under random-split evaluation, while also revealing substantial performance degradation for all classifiers under temporally ordered evaluation—a finding with direct implications for realistic AML deployment.