Experimental Analysis of Random Forest and LSTM Models for Optimizing CPU Efficiency in Cloud Computing

Cloud computing provides scalable and on-demand computational resources; nevertheless, effective utilization of the CPU is a critical issue because insufficient provisioning can result in higher latency and increased energy consumption. This research aims to measure, assess, and optimize CPU performance in cloud systems through the use of two machine learning models, namely, Random Forest (RF) and Long Short-Term Memory (LSTM), and subsequently using the Multi-Objective Grey Wolf Optimizer (MOGWO) to optimize the hyperparameters of the models to predict time series. The models were evaluated on the Google 2019 cluster trace using common metrics (Accuracy, Precision, Recall, and F1). The experimental findings demonstrate that optimization using MOGWO is an effective method for enhancing the performance of the models. Specifically, the accuracy of the Random Forest model increased from 65.15% to 89.70%, while the LSTM model improved from 95.02% to 98.60%, with consistent gains in precision, recall, and F1-score. These results verify the usefulness of the optimized models in forecasting the patterns of CPU usage and optimizing the use of CPU in cloud computing systems.