ARTIFICIAL INTELLIGENCE-BASED FRAUD DETECTION IN DIGITAL BANKING: A COMPARATIVE MACHINE LEARNING APPROACH ANOMALY DETECTION, ENSEMBLE ML BANKING, FINTECH, FRAUD PREVENTION

Abstract

The rapid growth of digital banking, mobile payments, internet banking, and fintech platforms has transformed the financial services industry while simultaneously increasing exposure to fraudulent activities. Traditional rule-based fraud detection systems often struggle to identify sophisticated, rapidly evolving, and previously unseen fraud patterns. Artificial Intelligence (AI) and Machine Learning (ML) provide promising solutions by enabling financial institutions to analyze large volumes of transactional data, detect anomalous behaviour, and continuously improve fraud identification. This research examines an AI-based approach to fraud detection in digital banking through a comparative analysis of machine learning techniques, with particular emphasis on anomaly detection and ensemble learning. Supervised algorithms such as Logistic Regression, Decision Tree, Random Forest, and Gradient Boosting are considered alongside anomaly-detection techniques such as Isolation Forest. The proposed methodology involves data preprocessing, feature engineering, class-imbalance handling, model training, and evaluation using accuracy, precision, recall, F1-score, Area Under the Receiver Operating Characteristic Curve (ROC-AUC), and fraud-detection rate. Because banking fraud datasets are typically highly imbalanced, the study emphasizes precision and recall rather than accuracy alone. The research proposes that ensemble machine learning models can provide more robust fraud detection than individual classifiers by combining multiple decision mechanisms and capturing complex nonlinear transaction patterns. The study further highlights the importance of explainability, real-time detection, data privacy, and adaptability in practical fintech environments. The proposed comparative framework can assist banks and fintech organizations in selecting appropriate AI techniques for reducing financial losses while maintaining legitimate customer transactions.