A method based on the Genetic Algorithm (GA) was developed to study the phase behavior of multicomponent and multiphase systems. Upon application of the GA to the thermodynamic models which are commonly used to study the VLE, VLLE and LLE phase equilibria, the physically meaningful values for the Binary Interaction Parameters (BIP) of the models were obtained. Using the method proposed in this work the activity coefficients for components at infinite dilution, obtained from the local composition based models, can be accurately predicted comparing to the experimental data available in the literature. In this work, a Global Optimization Procedure (GOP) based on the GA was developed to obtain the binary interaction parameters of the Wilson, NRTL and the UNIQUAC models for a number of systems at various temperatures. The VLE, VLLE and LLE values of the binary interaction parameters for the activity coefficients models were compared with those reported in the literature for the systems studied in this work. The results showed that the values reported for the binary interaction parameters can predict the activity coefficients at infinite dilutions for the components in the VLE, VLLE and LLE systems studied in this work. In order to confirm the accuracy of the results, the values for the activity coefficients at infinite dilution in binary solutions were compared with those reported in the literature. The comparison showed that the models studied using the proposed method can predict the physically meaningful binary interaction parameters among the species present in solutions. In addition to the accuracy, simplicity, generality and short CPU computation time of the proposed method should be mentioned as some of its clearest advantages.