Estimation of Heat of Formation for Chemical Systems using the Lasso Regression-Based Approach

Heat of formation (HoF) of a chemical species is one of the most essential thermodynamic properties to help understand and predict behaviors of a chemical system; however, it is very challenging to obtain accurate HoF values in large systems using traditional approaches, such as quantum mechanics. In this study, we propose a Lasso Regression-based machine learning approach, which is combined with the Reaction-based approach and Morgan fingerprints, to obtain reliable HoF values on-the-fly for an unknown chemical species. A dataset of species is taken into account for training and testing in order to evaluate the proposed machine learning approach, compared with the previous experimental results.