Abstract: In this article, neural networks (NN) with modified backpropagation (MBP) were employed for estimation and prediction of ¹³C NMR chemical shifts in alkanes from one or two through nine or ten carbon atoms. Carbon atoms in alkanes were determined by 16 descriptors which correspond to the so-called embedding frequecies of rooted subtrees or rooted subgraphs. These descriptors were equal to appearance numbers of smaller substructural skeletons composed of one through six carbon atoms (C1~C6). It was demonstrated that the employed descriptors offered a quite useful formal technique for the proper and adequate description of environment of carbon atoms in alkanes. Neural networks with different numbers of hidden neurons were examined. NN with three hidden neurons gave the best results. The results of NN computation were compared with those of multiple linear regression (MLR) calculations. Good results were obtained by both techniques.

Key words: Neural Networks in Spectroscopy, Chemical shifts of ¹³C NMR, Alkanes, Rooted subtrees and rooted subgraphs, Modified backpropagation (MBP), Multiple linear regression (MLR)