Predicting the thermodynamic stability of perovskite oxides using multiple machine learning techniques

The ultimate objective of material science and engineering is to design a material which gives good property. As part of reaching that aim, the material science community has been working hard to build massive data sets of material characteristics so that researchers can quickly get the properties of identified materials. The development of systems that utilize machine learning techniques to derive prediction models from current materials data is a current active field of research in material science. In this work, we used a dataset which has DFT-calculated perovskite oxide energies to create machine learning models to predict the thermodynamic phase stability of perovskite oxides. Convex hull analysis was used to determine phase stability, with Ehull (energy above the convex hull) providing a direct measure of the stability. Machine learning techniques like K-nearest Neighbor (K-NN), Decision Tree, Support Vector Machine (SVM), Random Forest, Neural Network (NN) and AdaBoost classifiers are used. Found that Out of all these classifiers AdaBoost works better.