The invention belongs to the technical field of material energetic analysis, and particularly relates to an energy prediction method and
system for a nano-hole-
alloy atomic cluster structure in
metal based on
machine learning, and the method comprises the steps: carrying out the energy prediction of a to-be-selected site on the surface of a nano-hole-
alloy atomic cluster; counting the number of vacancies and
alloy atoms in first and second adjacent shell
layers to form feature vectors; and inputting the
feature vector into a trained prediction model, outputting first
binding energy for describing the interaction between the new alloy atoms and the pure nano-pores and second
binding energy for describing the interaction between the new alloy atoms and the adsorbed alloy atoms, and summing to obtain total
binding energy. Training data is obtained through
density functional theory calculation, a gradient lifting
decision tree model is adopted, and features are
cut off from front ten neighbor shell
layers to front two neighbor shell
layers based on feature importance analysis. According to the method, the problems that the traditional DFT calculation amount is large and the
molecular dynamics precision is limited are solved, and efficient and accurate prediction of the binding energy of the
metal defect-solute
system is achieved.