The PREDICT (Predict pRoperties from Existing Database in Complex-alloys Territory) machine learning (ML) framework will take the input from a relaxed, 16 atom 2 x 2 x 1 unit cell binary FCC structure containing Ni, Cu, Au, Pd, and or Pt elements and predict the elastic constants, Young's moduli in various directions, bulk and shear moduli, and Poisson's ratio of the material of ternary, quaternary, and quinary structures in the Ni-Cu-Au-Pd-Pt system. This is acheived by using the bond lengths, the bond types, and the cohesive energy of the input structure as the descriptors in the model. The binary, ternary, quaternary, and quinary databases for this ML framework were obtained via density functional theory (DFT) simulations using the Vienna ab initio simulation package, using the stress-strain method as per Y. le Page and P. Saxe, “Symmetry-general least-squares extraction of elastic data for strained materials from ab initio calculations of stress,” Physical Review B - Condensed Matter and Materials Physics, vol. 65, no. 10, pp. 1–14, 2002, doi: 10.1103/PhysRevB.65.104104.

The mechanical properties predictions are achieved by reading in the data from the "binary_data" folder and splitting it into an 80-20 train-test split. The model that is used is a random forest regression model from the Scikit-Learn Python library. After the model is trained on the binary data, it can then be tested and used on the database of ternary strucutres found in the "ternary_data" folder, quaternary structures found in the "quaternary_data" folder, quinary structures found in the "quinary_data" folder, or a relaxed 2 x 2 x 1, 16 atom structure containing amounts of Ni, Cu, Au, Pd, and/or Pt of the users choice.

To reproduce the ML outputs from N. Linton and D.S. Aidhy, "A Machine Learning Framework for Elastic Constants Predictions in Multi-Principal Element Alloys," AIP Machine Learning, Under Review, one can also use the models that are saved in the "machine_learning" folder.