Supervised Machine Learning-Graph Theory Approach For Analyzing the Electronic Properties of Alkanes

Abstract

The combination of advanced scientific computing and quantum chemistry improves the existing approach in all chemistry and material science fields. Machine learning has revolutionized numerous disciplines within chemistry and material science. In this study, we present a supervised learning model for predicting the HOMO and LUMO energies of alkanes, which is trained on a database of molecular topological indices. We introduce a new moment topology approach has been introduced as molecular descriptors. Supervised learning utilizes artificial neural networks and support vector machines, taking advantage of the correlation between the molecular descriptors. The result demonstrate that this supervised learning model outperforms other models in predicting the HOMO and LUMO energies of alkanes. Additionally, we emphasize the importance of selecting appropriate descriptors and learning systems, as they play crucial role in accurately modeling molecules with topological orbitals.