Abstract: The neural network was employed to study the phase equilibrium data of the aqueous two-phase system (pH=5,6,7 and 8), which consisted ofC₄mimBr and K₃C₆H₅O₇/H₃C₆H₅O₇. The method of two dimensional optimization was proposed in order to obtain the appropriate node number of the hidden layer and the error control value. The accuracy comparison between the suggested model and the thermodynamic equations showed that the neural network was almost equal to Wilson equation, and superior to NRTL equation.