Magnetic topological materials (MTMs) hold significant potential for advancements in spintronic technologies; however, their application is hindered by low Curie temperatures and limited experimental data. In this work, we establish a machine learning pipeline aimed at predicting the Curie temperature of MTMs by aggregating magnetic and structural information from various databases, and subsequently applying the trained model to a set of topological materials. A meticulously curated dataset comprising 7365 distinct materials was processed and featurized using Matminer and Pymatgen , and subsequently trained with Random Forest and XGBoost regressors, resulting in R 2 scores of 0.864 and 0.876, respectively. Predictions for 200 MTMs indicated a strong correlation between the models (r = 0.94) and demonstrated high accuracy on a validation subset of 26 materials (Pearson r > 0.95). We emphasize 20 high T C candidates for DFT and experimental validation, showcasing the efficacy of machine learning driven screening in the discovery of new materials and predicting their properties. • ML pipeline using magnetic materials data to predict Tc of Topological Materials. • The XGB showed R 2 score 0.8763 and RF performed to the R 2 score of 0.8637. • Predicted and tabulated T C of 200 different magnetic topological materials. • Validated 26 materials with Pearson Correlation coefficient > 0.95. • The training data was very sparse, many columns had up to 70% null values.
Rehman et al. (Wed,) studied this question.