Published Paper Details:

This paper explores the development and implementation of AI-based workload forecasting techniques to address the complex, dynamic demands faced by e-commerce cloud platforms. With the rapid growth of online retail and the accompanying fluctuations in user activity, traditional static and statistical forecasting methods often fail to provide the accuracy and responsiveness required for optimal cloud resource management. This research investigates advanced machine learning models, including LSTM, GRU, and Transformer architectures, and benchmarks their performance against conventional time series approaches like ARIMA and Prophet. Using real-world e-commerce workload data, the study demonstrates that deep learning models significantly enhance forecast precision, especially during peak demand periods driven by promotional events and shifting consumer behavior. A pilot deployment further validates the models' practical impact on dynamic resource scaling, cost efficiency, and service reliability. By integrating explainable AI techniques, the paper also addresses the need for interpretability and stakeholder trust in automated forecasting systems. The findings highlight both the transformative potential and the operational challenges of adopting AI for workload prediction, offering actionable insights for e-commerce businesses and cloud providers aiming to build more intelligent, resilient, and sustainable digital infrastructures.