Forecasting Cloud Storage Costs using Machine Learning on usage Patterns
Naga Ravi Teja Vadrevu, Vivek Bagmar, Paras Doshi, Shailendra Shrivastava · 2025
This research examines strong techniques to estimate cloud storage costs using utilization pattern analysis and core operational metrics. With businesses embracing cloud platforms at a fast-growing rate, the complexity and dynamics of storage pricing models pose serious financial challenges. This research discusses the increasing demand for effective forecasting tools that account for real-time usage parameters, operational trends, and configuration parameters. Current pricing calculators only provide an approximate prediction, taking into account little or no operational dynamics and temporal behavior. This research develops a machine learningdriven solution to fill this gap, presenting predictive insights that can help IT budgeting, resource planning, and cost optimization. The research utilizes various machine learning algorithms to discover the most contributing drivers of cost variation in different storage scenarios. Applying a synthetically created dataset of$\mathbf{5, 0 0 0}$records and$\mathbf{1 6}$unique features-such as data volume, region, storage tier, and operation counts-the study finds that data egress volume and total storage size are the strongest predictors of cost variation. Among the models that were tested, XGBoost had the highest predictive accuracy of 98.74%, surpassing other models like Linear Regression (90.51%), Support Vector Regression (91.67%), and Gradient Boosting (98.23%). These findings highlight the importance of advanced predictive modeling in predicting cloud spending. The research finally offers real-world insights into the application of data-driven methods for cost management in enterprise cloud systems, enabling organizations to enhance cost efficiency and strategic decision-making.