Machine Learning-Driven Density Prediction for Nanomaterials DOI Creative Commons

Shams Ansaf

Wasit Journal of Pure sciences, Год журнала: 2024, Номер 3(4), С. 281 - 288

Опубликована: Дек. 30, 2024

This paper provides a machine learning method that uses band gap and chemical composition data from the Materials Project database to predict density of nanomaterials. We developed an improved Random Forest Regressor compared it against several baseline models, including Linear regression, demonstrate superior performance our approach Using rigorous preprocessing procedure, we combined elemental characteristics extracted formulae with data. To improve random forest hyperparameters boost predictive power model, employed grid search cross-validation. Key components have biggest effects on nanomaterial were identified via feature importance analysis. Insights into structure-property correlations in nanomaterials gained by examining link between gap. Because allows for quick estimates, this work shows how can speed up discovery design By enabling high-throughput screening directing experimental efforts materials synthesis characterization, created model be useful tool nanotechnology researchers engineers.

Язык: Английский

Applications of density functional theory and machine learning in nanomaterials: A review DOI

Nangamso Nathaniel Nyangiwe

Next Materials, Год журнала: 2025, Номер 8, С. 100683 - 100683

Опубликована: Апрель 28, 2025

Язык: Английский

Процитировано

0
DFT-PBE band gap correction using machine learning with a reduced set of features DOI
Ibnu Jihad, Miftah Hadi S. Anfa, Saad M. Alqahtani

и другие.

Computational Materials Science, Год журнала: 2024, Номер 244, С. 113153 - 113153

Опубликована: Июнь 15, 2024

Язык: Английский

Процитировано

3
Explainable artificial intelligence for machine learning prediction of bandgap energies DOI

Taichi Masuda,

Katsuaki Tanabe

Journal of Applied Physics, Год журнала: 2024, Номер 136(17)

Опубликована: Ноя. 4, 2024

The bandgap is an inherent property of semiconductors and insulators, significantly influencing their electrical optical characteristics. However, theoretical calculations using the density functional theory (DFT) are time-consuming underestimate bandgaps. Machine learning offers a promising approach for predicting bandgaps with high precision throughput, but its models face difficulty being hard to interpret. Hence, application explainable artificial intelligence techniques prediction necessary enhance model's explainability. In our study, we analyzed support vector regression, gradient boosting random forest regression reproducing experimental DFT permutation feature importance (PFI), partial dependence plot (PDP), individual conditional expectation plot, accumulated local effects plot. Through PFI, identified that average number electrons forming covalent bonds mass elements within compounds particularly important features models. Furthermore, PDP visualized dependency relationship between characteristics constituent bandgap. Particularly, revealed there where decreases as increases. This result was then theoretically interpreted based on atomic structure. These findings provide crucial guidance selecting descriptors in developing high-precision this research demonstrates utility methods efficient exploration potential inorganic semiconductor materials.

Язык: Английский

Процитировано

0
Machine Learning-Driven Density Prediction for Nanomaterials DOI Creative Commons

Shams Ansaf

Wasit Journal of Pure sciences, Год журнала: 2024, Номер 3(4), С. 281 - 288

Опубликована: Дек. 30, 2024

This paper provides a machine learning method that uses band gap and chemical composition data from the Materials Project database to predict density of nanomaterials. We developed an improved Random Forest Regressor compared it against several baseline models, including Linear regression, demonstrate superior performance our approach Using rigorous preprocessing procedure, we combined elemental characteristics extracted formulae with data. To improve random forest hyperparameters boost predictive power model, employed grid search cross-validation. Key components have biggest effects on nanomaterial were identified via feature importance analysis. Insights into structure-property correlations in nanomaterials gained by examining link between gap. Because allows for quick estimates, this work shows how can speed up discovery design By enabling high-throughput screening directing experimental efforts materials synthesis characterization, created model be useful tool nanotechnology researchers engineers.

Язык: Английский

Процитировано

0