A reaction-based ensemble machine learning method of determining the heat of formation in chemical systems

Abstract

For a species, Heat of Formation is one of the most critical properties to get a deeper understanding of the behavior of the chemical system. So far, chemists have spent their time to find the suitable approaches by using complex calculations and lots of experiments; therefore, most common methodologies have achieved the HoF value in various levels of accuracy. However, it is still a problem to obtain the correct HoF value. Understanding those issues, in this proposed solution, the Lasso Regression approach will use to obtain the HoF value of an unknown species. This approach will take the usage of Morgan Fingerprint to describe the considered species and its following reactions, along with the calculated values. Finally, the result will be compared with the previous approach to observe the differences.