There is growing interest in circumventing quantum chemical calculations for material screening purposes. Recently many statistical methods were applied to perform such calculations. However many of such methods either apply to a limited subset of structures or use non-ab initio features for calculation. Here we introduce a method of bootstrapping along with various machine learning algorithms to model energy-geometry dependency of various chemical structures including small organic and inorganic molecules, and ions. Our method reproduced quantum calculation results in a fraction of its time and within about 3% MAPE and a R2=0.99.

General machine learning based predictive method for energy-geometry dependency of molecules

256th National Meeting (2018)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.