Data Center Predictions using MATLAB Machine Learning Toolbox

Moises Levy, Daniel Raviv, Justin Baker

Abstract: This paper describes a predictive model using MATLAB machine learning tools for nonlinear regression, support vector machine and neural network. Data center performance measures such as run time, waiting time and energy consumption, are predicted based on past events, total workload input and number of IT resources. Predictive analytics enables evidence-based decision making in the case of uncertainty and risk, and helps to evaluate areas of improvement.