Chemistry Data Mining to Discover Molecular Interactions

Sammanfattning: In the computational chemistry field, the area of energy functions has not seen any changes for decades, since the functions which were first created still work quite well today. In this project we test to see if we can make improvements to those energy functions by doing data mining on a dataset of off-equilibrium molecule states, with the objective of being able to predict the energy of a molecule just by knowing its structure. We tested what results we could obtain by adding different levels of complexity to represent the structure which describes the molecule and see what contributions each level has to the final energy of the molecule. This was examined for both elements and atom types. As a conclusion, we obtained an energy function which potentially could predict the energies, but further work has to be done to improve this function.