Hybrid Machine Learning Models for Fraud Detection in Cloud-Enabled Supply Chains

Abstract

The increasing digitalization of supply chains, accelerated by cloud adoption, has introduced new opportunities for efficiency alongside heightened risks of fraudulent activities. Conventional fraud detection techniques often fail to adapt to the dynamic, high-volume, and heterogeneous nature of cloud-enabled transactions. This paper proposes a hybrid machine learning framework that integrates supervised and unsupervised algorithms to enhance fraud detection accuracy within cloud-based supply chain systems. The approach leverages supervised classifiers for recognizing known fraud patterns while employing unsupervised anomaly detection to capture novel or evolving threats. By combining these techniques, the model addresses the limitations of single-method approaches, such as overfitting and limited generalization. Empirical evaluation on simulated supply chain datasets demonstrates improved detection rates, reduced false positives, and robust scalability in cloud environments. The study contributes to the growing body of knowledge on intelligent fraud management by highlighting the value of hybrid architectures for safeguarding digital supply chains. Future research directions include extending the model to multi-cloud ecosystems and incorporating real-time learning capabilities for adaptive risk management.