EVOLUTIONARY INSIGHTS INTO ENSEMBLE LEARNING MODELS FOR ADVANCED ONLINE FRAUD DETECTION

Abstract

One of the largest existing problems in the financial industry is “Credit Card Fraud” (CCF) as it incurs significant financial losses and erodes consumer confidence. Accordingly, this paper aims at describing how the future online fraud detection system particularly in online “Credit Card Fraud Detection” (CCFD) which incorporates “Deep Learning” (DL), “Machine Learning” (ML), and “Ensemble Learning” (EL) models. The algorithms assessed include: “Logistic Regression”, “Random Forest” and “Decision Trees”, due to their capability of handling big data and model interpretability. Consequently, the DL approaches, including CNNs, LSTM networks, and GANs, are suggested to exhibit better results in the recognition of complex fraud features. Hence, the performance of various EL approaches such as soft voting, hybrid, and weighted methods are investigated to understand their suitability for model fusion. Also, this review offers an understanding of feature engineering, data preprocessing, and anomaly detection, all of which are essential in enhancing the performance of fraud detection systems. Therefore, based on the recent techniques of ML, DL, and EL, the study aims at providing actionable insights to the researchers and practitioners regarding the existing fraud detection strategies. These results further vindicate the need to recalibrate models and integrating human factor in combating the intricate strategies of the defrauders in enhanced measures of preventing monetary loss.