Application of Online Computational Platform of Materials Data Mining (OCPMDM) in Search for ABO3 Perovskites with Multi-Properties

Abstract

OCPMDM is an online platform specially developed for researchers who do not have any programming basics to perform material machine learning projects, in which the processing of ABO3 perovskite machine learning has even reached automation. In this work, we used OCPMDM to discover perovskite materials with multi-properties to demonstrate some functions of the platform, including the descriptor filling, regression, classification, pattern recognition, and virtual screening. The results of LOOCV and independent test of the constructed regression and classification models for Curie temperature and band gap show the reliable predictive ability of the models via the platform. In the pattern recognition optimization area, the occupancy rate of superior samples with high Curie temperature and suitable band gap reached 92.73% and 80%, respectively. In addition, we also screened out 8 candidates with higher Curie temperature and proper band gap for experiments.