A model for predicting resources for on-premise applications

Abstract

Accurate prediction of resources is a cliallenging problem in any environment. Effective provisioning of resources for on-premise applications with varied performance requirements requires an accurate prediction of resources. Towards this objective, a prediction model, namely. Multilayer Perceptron has been proposed in this work. The prediction model is trained using a dataset generated from TPC-W benchmark based online application and tested for new requirements. Its prediction accuracy has been compared with that of two other prediction models such as Linear Regression and Support Vector Regression. The Multilayer perceptron model is found to exhibit a better accuracy of 91.8 percentage.