AI-Powered Big Data Analytics for Scalable Cloud and Edge Computing

The emergence of rapid growth in digital financial transactions has increased the risk of fraud, which requires scalable and intelligent fraud detection paradigms. Existing rule-based systems as well as cloud-centric architectures are incapable of achieving the necessary trade-off among detection accuracy, latency, and resource consumption. In this paper, we introduce an AI-driven big data analytics paradigm using hybrid cloud–edge architecture to detect fraud in real time. Big financial transaction data is harvested, preprocessed, and utilized to train sophisticated machine learning and deep learning models in the cloud and deploy lightweight versions on edge devices like ATMs and mobile banking apps for low-latency inference. The architecture combines up-to-date models, such as Random Forests, CNNs, Transformers, and a new Hybrid Model, that are optimized for high-dimensional and imbalanced data. Experimental results on the IEEE-CIS Fraud Detection dataset show that the Hybrid Model performs better, with an accuracy of 97%, precision of 0.88, recall of 0.85, and an F1-score of 0.86 compared to baselines. Confusion matrix and ROC curve (AUC = 0.98) further support the model to reduce both false positives and false negatives. Through the integration of cloud-based retraining with edge-powered inference, the presented framework minimizes bandwidth usage, decreases operational expenses, and improves real-time decision-making. These results identify the promise of AI-empowered cloud–edge synergy as a scalable approach for financial fraud detection across contemporary digital environments.