For predicting the standard entropy of a binary inorganic compound, two novel connectivity indexes mQ,mG and their converse indexes mQ',mG' based on adjacency matrix of molecular graphs and ionic parameters gi, qi were pro-posed. The qi and gi are defined as qi=(1.1+Zi1.1)/(1.7+ni), gi:(1.4d-Zi)/(0.9+ri+ri-1), where Zi, ni, ri are the charge numbers, the outer electronic shell primary quantum numbers, and the radii of ionic I respectively. The good Quantitative Structure-Property Relationship (QSPR) models for the standard entropies of binary inorganic com-pound can be constructed from 0Q,0Q',1G, and 1G', by using a multivariate linear regression (MLR) method and an artificial neural network (NN) method. The correlation coefficient r, the standard error s, and the average absolute deviation of the MLR model and the NN model are 0.9905, 8.29 J·K-1,mol-1 and 6.48 J·K-1·mol-1, and 0.9960,5.37 J·K-1·mol-1 and 3.90 J·K-1·mol-1, respectively, for 371 binary inorganic compounds (training set). The cross-validation by using the leave-one-out method demonstrates that the MLR model is highly reliable from the point of view of statistics. The correlation coefficients, standard deviations and average absolute deviations of pre-dicted values of the standard entropies of other 185 binary inorganic compounds (test set) are 0.9897, 8.64 J·K-1·mol-1 and 6.84 J·K-1·mol-1, and 0.9957, 5.63 J·K-1·mol-1 and 4.18 J·K-1·mol-1 for the MLR model and the Nnmodel, respectively. The results show that the current method is more effective than literature methods for estimat-ing the standard entropy of a binary inorganic compound. Both MLR and NN methods can provide acceptable mod-els for the prediction of the standard entropies of binary inorganic compounds. The NN model for the standard en-tropies appears to be more reliable than the MLR model.
展开▼