[16p-D209-1~14] 11.1 Fundamental properties

[16p-D209-12] Effects of datasets and features on superconducting Tc prediction using machine learning

〇Kaname Matsumoto¹, Tomoya Horide¹, Masaki Mito¹ (1.Kyshu Inst. Technol.)

Keywords:superconductor, critical temperature, machine learning

So far, we have discovered a new superconducting material, AlTixOy, which is expected to have a Tc of over 55 K by using a high-pressure torsion method and machine learning (ML). In this study, we were able to predict high-Tc substance that was consistent with the experiment by ML, but in order to increase the versatility of the prediction performance, human intervention is required at this time, such as including non-superconducting substances in the dataset, rather than relying solely on ML techniques. It is important to devise new methods or conduct new experiments to expand the range of prediction. The talk will discuss these points.