Big Data-Driven Artificial Intelligence (AI) Models for Predictive Fraud Detection in Banking Networks

The increasing cases of digital banking transactions by leaps and bounds have contributed to increasing the difficulty of discovering the fraud that are concealed within the immense, massive, and incredibly uneven datasets. This research delves into the topic of financial network fraud prediction using AI models powered by big data, specifically ML and DL. The Credit Card Fraud Detection dataset is utilized, which mimics actual financial data and undergoes extensive preprocessing including cleansing, feature extraction, normalization, and class balancing using SMOTE. Two models are put forward, including XGBoost to tackle structured tabular pattern recognition and a CNN architecture to extract deep features. The accuracy, precision, recall, F1-score, and ROC curve are measures used to evaluate both models. Experimental results indicate that XGBoost was accurate at 99.98 percent, which is better than other ML models, whereas CNN was accurate at 99.97 percent meaning that there is superb performance at identifying nonlinear fraud patterns. Comparative analysis proves that the two models are complementary to each other where XGBoost will ensure interpretability and computational efficiency whereas CNN will address other intricate data dependencies. Future research may integrate hybrid architectures combining ML-DL features with real-time data streams for adaptive fraud prevention in large-scale financial systems.