Ensemble Learning Framework for Financial Fraud Detection Using Temporal Transaction Pattern Analytics

Abstract

Financial fraud has increased with the rapid growth of digital payment platforms, making accurate fraud detection a major challenge. Traditional machine learning (ML) methods often struggle to adapt to evolving fraud patterns, effectively capture temporal transaction behaviors, and accurately distinguish fraudulent transactions. To address these limitations, research proposes an Ensemble Learning Framework (ELF) for intelligent financial fraud detection using temporal transaction pattern analytics. The framework utilizes a financial fraud transaction analytics dataset that includes 15,000 instances of financial transactions and 44 features of temporal transactions comprising time-stamped transaction records, customer information, transaction attributes, and fraud labels. Initially, the preprocessing stage employs K-Nearest Neighbors (KNN) imputation for missing value handling, followed by Isolation Forest and Local Outlier Factor (LOF) for outlier detection. Subsequently, Neighborhood Components Analysis (NCA) is employed to extract highly discriminative transaction features while preserving class separability. Finally, the framework integrates the proposed Dynamic Artificial Rabbits Optimization-Driven Intelligent Light Gradient Boosting Machine (DARO-ILightGBM), where the Dynamic Artificial Rabbits Optimization algorithm performs adaptive hyperparameter optimization, while the Intelligent Light Gradient Boosting Machine Model accurately classifies fraudulent transactions. Experimental evaluation demonstrates that the proposed framework achieves a superior Receiver Operating Characteristic–Area under the Curve (ROC-AUC) of 0.9765, Precision–Recall Area under the Curve (PR-AUC) of 0.9677, F1-score of 0.9720, Precision of 0.9740, and Recall of 0.9685using Python 3.11. Overall, the proposed ELF effectively captures temporal transaction patterns and provides a scalable solution for intelligent financial fraud detection in modern digital banking environments.