Optimizing E-Commerce Fund Transfers with Gradient Boosted Decision Trees (GBDT), Markov Decision Processes (MDPs), and Serverless Computing for Biometric Security

Abstract

Background:
E-commerce platforms struggle to maintain transaction efficiency, optimise fund transfers, and guarantee security. As fraud risks and transaction complexity rise, these problems become more complicated and call for creative solutions.
Objectives:

With an emphasis on biometric security, this work attempts to improve fund transfer optimisation through the use of serverless computing, Markov Decision Processes (MDPs), and Gradient Boosted Decision Trees (GBDT).
Methods:
We integrate biometric security for authentication, serverless computing for efficiency, MDPs for decision optimisation, and GBDT for fraud detection. Metrics including accuracy, AUC, and latency are used to assess the system's performance.
Empirical Results:
93% accuracy, 0.95 AUC, and 1 ms latency were attained by the entire technique. It performed better than other settings in terms of security, fraud detection, and real-time transaction optimisation.
Conclusion:
By increasing security, efficiency, and fraud detection, the suggested approach greatly improves e-commerce fund transfer optimisation, showcasing its promise for real-time, scaled transaction systems.