NEURAL NETWORKS AND GRAPH-THEORY AS COMPUTATIONAL TOOLS FOR PREDICTING POLYMER PROPERTIES

A new computational methodology is presented for making rapid and accurate predictions of chemical, physical and mechanical properties of polymers based on their molecular structure. The method uses a set of topological indices derived from graph theory to numerically describe the structure of a monomeric repeating unit for a given polymer (structural descriptors) and relates these indices to a set of polymer properties by utilizing an artificial neural network. The neural network is able to efficiently formulate all of the correlations (i. e., between structural descriptor-property, property-property, structural descriptor-structural descriptor: both linear and nonlinear dependencies) necessary to make accurate predictions. Results have been obtained for up to 9 properties of 357 different polymers with an average prediction error of < 3% and a maximum error of 12%, demonstrating superiority over other quantitative structure/property relationships for polymers.