A Preliminary Neural\nNetwork-Based Composite Method\nfor Accurate Prediction of Enthalpies of Formation

A composite method, named ANN-G3S, is introduced, adapting\nfrom\nG3S theory and employing distinct sets of multiplicative scale factors.\nAn artificial neural network (ANN)-based classification model is utilized\nto select optimal sets of four scale factors for electronic correlation\nand basis set expansion terms in electronic systems. The correlation\nand basis set terms are scaled by four parameters, two for atoms and\nthe other two for molecules. The ANN model is trained on the G3/05\ntest set to identify the best parameter set for each electronic system.\nTo validate the method, 10% of the structures from the test set are\nrandomly excluded from training and optimization, forming a separate\nvalidation set. The method demonstrates a mean deviation of 1.11 kcal\nmol<sup>–1</sup> for the G3/05 set and 0.89 kcal mol<sup>–1</sup> for the validation set, close to the value presented by the G4 method\nand surpassing the accuracy of the G3 method of 1.19 kcal mol<sup>–1</sup> with significantly reduced computational cost. This\nmethod shows advantages by eliminating the need for purely empirical\ncorrections, thereby enhancing both efficiency and accuracy in predicting\nheats of formation.