Hybrid Approach for Virtual Machine Utilization Prediction in Cloud Environment

Predicting the virtual machine utilization accurately has become the key aspect in enhancing overall performance and efficiency of computer system along with optimizing the resource allocation in cloud computing environment. The research paper formulates amalgamation of various machine learning models with an ensemble averaging technique to forecast VM utilization. Numerous ML models have been considered to attain accuracy in VM utilization prediction namely, k-Nearest Neighbors, Feed-Forward Neural Networks and Recurrent Neural Networks with Long Short-Term Memory. The considered models are competent in apprehending varying range of underlying patterns and correlations existing in data. The ensemble averaging tends to yield the prediction amalgamation while making efficient use of collaborative knowledge attained from all the considered models. The thorough assessment of developed approach has been done by utilizing the dataset comprising actual VM consumption details. Real-world VM utilization datasets have been used to evaluate the proposed method, and wide range of performance measures, such as Mean Squared Error, Mean Absolute Error, and R-squared score, proved the efficacy of ensemble in terms of enhanced forecast accuracy. The simulation results claimed that the ensemble technique outperformed individual models by yielding better outcomes; henceforth, offering reliable solution for VM usage prediction in dynamic computing environment.