Classification of magnetic ground states and prediction of magnetic moments of inorganic magnetic materials based on machine learning

Magnetic materials are important basic in the information age. Different magnetic ground states prerequisite for wide application of materials, among which ferromagnetic state is a key requirement future high-performance materials. In this paper, machine learning used to study classification ferromagnetic, antiferromagnetic, ferrimagnetic and paramagnetic inorganic prediction moments We obtain 98888 data from Materials Project database, containing material ids, chemical formulae, CIF files, moments, extract 582 elemental structural features by using Matminer. design two-step feature selection method. first step, RFECV evaluate one remove redundant without degrading model accuracy. second we rank further refine select most model, 20 selected respectively. Among features, it found that electronegativity, atomic own moment number unfilled electrons peripheral orbitals all make contributions moments. build random forest, quantitatively models 10-fold cross-validation approach, results show constructed has sufficient accuracy generalization capability. test set, an 85.23%, precision 85.18%, recall 85.04%, F1 score 85.24%; goodness-of-fit 91.58% average absolute error 0.098 μ<sub>B</sub> per atom. This provides new method choice high-throughput screening predicting