Prediction of Space Groups for Perovskite-Like A2BB'O6 Compounds
The prediction of new compounds having such composition as BIIIB'VO6 was carried out, the
type of distortion of their perovskite-like lattice and the space group were predicted, and the crystal lattice
parameters of the predicted compounds were estimated. For the prediction, only the property values of the
chemical elements were used. The programs based on machine learning algorithms for different variants of
neural networks, a linear machine, the formation of logical regularities, k-nearest neighbors, and support
vector machine showed the best results when predicting the type of distortion of a perovskite-like lattice.
When evaluating the lattice parameters, programs based on algorithms for orthogonal matching pursuit and
automatic relevance determination regression were the most accurate methods. The prediction accuracy for
the type of distortion of perovskite-like lattice was no less than 74%. The accuracy of estimating the lattice
linear parameters was within ±0.0120–0.8264 Å, and the accuracy of angles β for the monoclinic distortion
of the lattice amounted to ±0.08°–0.74°. The calculations were carried out using systems based on machine
learning methods. To evaluate the prediction accuracy, an examination recognition in the cross-validation
mode was used for the compounds included in the sample for machine learning. The predicted compounds
are promising for searching for novel magnetic, thermoelectric, and dielectric materials.