Document Type

Article

Rights

This item is available under a Creative Commons License for non-commercial use only

Disciplines

1.2 COMPUTER AND INFORMATION SCIENCE, Information Science

Publication Details

Molecules 2021,26, 8.

https://dx.doi.org/10.3390/molecules26010008

Abstract

n this study, we have investigated quantitative relationships between critical temperaturesof superconductive inorganic materials and the basic physicochemical attributes of these materials(also called quantitative structure-property relationships). We demonstrated that one of the mostrecent studies (titled "A data-driven statistical model for predicting the critical temperature of asuperconductor” and published in Computational Materials Science by K. Hamidieh in 2018) reportson models that were based on the dataset that contains 27% of duplicate entries. We aimed todeliver stable models for a properly cleaned dataset using the same modeling techniques (multiplelinear regression, MLR, and gradient boosting decision trees, XGBoost). The predictive ability ofour best XGBoost model (R2 = 0.924, RMSE = 9.336 using 10-fold cross-validation) is comparableto the XGBoost model by the author of the initial dataset (R2 = 0.920 and RMSE = 9.5 K in ten-foldcross-validation). At the same time, our best model is based on less sophisticated parameters, whichallows one to make more accurate interpretations while maintaining a generalizable model. Inparticular, we found that the highest relative influence is attributed to variables that represent thethermal conductivity of materials. In addition to MLR and XGBoost, we explored the potential ofother machine learning techniques (NN, neural networks and RF, random forests

DOI

https://dx.doi.org/10.3390/molecules26010008


Share

COinS