Machine Learning Prediction of magnetic properties for Fe-based and Pt-based alloys

There are two approaches to getting the Curie/Néel temperature: the classical one is to prepare an alloy of a certain composition and measure its temperature by experimental methods. In contrast, another approach is to predict the magnetic properties ahead of the experiment based on the physical and chemical properties of the material via machine learning methods.In this paper, we take the second approach and apply machine learning techniques to predict the Curie/Néel temperature for Fe-based and Pt-based alloys. As control factors, we chose atomic order and doping the binary alloy with a third element. This work applies the earlier proposed technique to predict the Curie temperature in binary alloys [1] and extends it to applications in three-component compounds, using the Fe-Pt-Pd alloy as a modeling system. The analysis shows that the Voting Ensemble is the best performing model that achieves a Normalized Root Mean Square Error (NRMSE) of 0.051.