Microstructural Feature Extraction And Analysis Of Multiple Linear Regression

The relationship between material structure and properties has always been the focus of research in the field of material calculation.The microstructure of materials determines the properties of electrical properties.Through the extraction and characterization of microstructure features to quantitatively analyze the composition,structure and properties of materials,this process has traditionally involved long and expensive trial and error experiments.Consider using computational software and machine learning techniques for numerical simulation to achieve an effective combination of experiments,calculations,and data.Graphene is a semiconductor with zero band gap,and its microstructural characteristics such as size and edge configuration are important factors in determining its metallicity and quantum mechanics.By extracting the microstructure by machine learning method,the material with the target attribute can be prepared in a targeted manner,and the structureperformance relationship can be analyzed more clearly,and the general experimental conclusion can be verified.In this paper,multivariate linear regression algorithm is used to extract the microstructure characteristics of materials based on density functional theory.The main geometrical features of graphene nanostructures and their influence on electronic properties are used for pretreatment analysis.The multivariate statistical method is combined with the self-consistent density charge density functional tightbinding algorithm to perform multivariate linear regression training and prediction of material microstructure characteristics.The univariate analysis method was used to infer the effects of several characteristic variables on the electron band gap and Fermi level energy,and the correlation results were visualized in the form of scatter plots.How to affect the electronic properties of graphene nanosheets is analyzed for the microstructure surface area,edge type and aspect ratio involved.The electronic properties of the extracted graphene structure are linked with the microstructure characteristics,and the model training and prediction are carried out.Several characteristics reflecting the electron band gap and Fermi level energy are extracted by the univariate analysis method.