AI-Based Predictive Cloud Load Management System

Cloud computing environments experience highly dynamic workloads that often lead to inefficient resource utilization when managed using traditional reactive scaling strategies.Reactive mechanisms trigger scaling operations only after system resources are already under heavy load, resulting in performance degradation and increased response latency.This paper proposes an AI-Based Predictive Cloud Load Management System designed to forecast server load patterns and recommend proactive scaling decisions before resource saturation occurs.The proposed system employs a Random Forest Regressor trained on historical CPU utilization datasets to predict upcoming load levels based on time-slot features.The predicted load values are evaluated using predefined thresholds to determine scaling actions such as scaling up or scaling down virtual machine resources.The architecture consists of a React-based monitoring dashboard, a Flask backend for business logic and machine learning inference, and a MongoDB database for storing prediction logs and system metrics.Experimental analysis demonstrates that the model successfully identifies peak usage periods with high accuracy and generates effective scaling recommendations.The system helps administrators maintain service availability, improve infrastructure utilization, and significantly reduce operational costs by enabling proactive cloud resource management.