Machine learning based prediction of metal hydrides for hydrogen storage, part I: Prediction of hydrogen weight percent

Abstract

The openly available database provided by the US Department of Energy on hydrides for hydrogen storage were analyzed through supervised machine learning to rank features in terms of their importance for determining hydrogen storage capacity referred to as hydrogen weight percent. In this part: we employed four models, namely linear regression, neural network, Bayesian linear regression and boosted decision tree to predict the hydrogen weight percent. For each algorithm, the scored labels were compared to the actual values of hydrogen weight percent. Our investigation showed that boosted decision tree regression performed better than the other algorithms achieving a coefficient of determination of 0.83.