Abstract: A back propagation neural network model with one hidden layer was established to predict the liquidliquid equilibrium data of n-heptane + toluene + (sulfolane and N-methylpyrrolidone) system under atmospheric pressure. The equilibrium data of 303.15 K and 313.15 K were employed to train the network, and then the optimized model parameters were obtained. The equilibrium data of 323.15 K were used to validate its predicting capability. The results showed that average absolute deviations of the correlated (training) mole fractions from experimental data (303.15 K and 313.15 K) were 0.87% and 0.71%,respectively, and that of the predicted mole fractions from experimental data (323.15K) was 1.59%.