AI-Driven Cost Optimization and Security Enhancement in AWS

Yunus israr, Abhijeet Singh, Manish Kumar

Keywords: AWS cost prediction, machine learning, service optimization, cloud security, XGBoost, cloud cost management.

Abstract: Cloud services have become ingrained in today's businesses with Amazon Web Services (AWS) at the top of the services list. However, the nature of the pricing model has added considerable complexity to their usage in an already complex cloud service. We introduce a machine learning-based methodology to predict cost and optimize AWS services. Our approach to predicting monthly AWS costs includes multiple regression models: Random Forest, Extreme Gradient Boosting, and long short term memory (LSTM) networks. Furthermore, our approach enables cost optimization and anomaly detection, increasing delivery efficiency and security. The experimental results show that XGBoost yields the most accurate forecast with RMSE at 3.92. This process allows organizations to predict their costs and reduce wasteful expenses while improving overall resource efficiency. Moving forward, we plan to implement our methodology in real-time using AWS Lambda and SageMaker.

Primary Area: unsupervised, self-supervised, semi-supervised, and supervised representation learning

Submission Number: 23670