Enhanced Feature Engineering and Algorithm Optimization for Real-Time Detection of Synthetic Identity Fraud and Money Laundering in Financial Transactions

The proliferation of digital financial transactions has created unprecedented opportunities for sophisticated fraud schemes, particularly synthetic identity fraud and money laundering activities that evade traditional rule-based detection mechanisms. This research introduces an enhanced feature engineering framework coupled with optimized machine learning algorithms to address the dual challenges of improving detection accuracy while minimizing false positive rates. The proposed methodology integrates temporal, behavioral, and network-based features specifically designed to capture the subtle patterns characteristic of synthetic identity fraud and money laundering transactions. Seven (including stacking ensemble) machine learning algorithms were systematically evaluated using real-world financial transaction datasets, with comprehensive performance analysis conducted through stratified cross-validation. Experimental results demonstrate that XGBoost achieved an F1-score of 0.938 and a Precision of 0.947, delivering the best balance between accuracy and real-time performance.