Predicting new superconductors and their critical temperatures using unsupervised machine learning

We used the superconductors in the SuperCon database to construct element vectors and then perform unsupervised learning of their critical temperatures (Tc). Only the chemical composition of superconductors was used in this procedure. No physical predictors (neither experimental nor computational) of any kind were used. We achieved the coefficient of determination R ' 0.93, which is comparable and in some cases higher then similar estimates using other artificial intelligence techniques. Based on this machine learning model, we predicted several new superconductors with high critical temperatures. We also discuss the factors that limit the learning process and suggest possible ways to overcome them.