In the present paper, to determine the pressure-dependent yield surface of polypropylene/nanoclay nanocomposites, the extended Drucker-Prager yield criterion is used and its parameters are derived by a combined experimental/numerical/optimization approach. In this method, the difference between the experimental and numerical results obtained from three-point bending test is minimized. In order to alleviate the burdensome numerical simulation, a surrogate model based on Kriging method is used to estimate the cost function. The optimum of this function is obtained by maximizing expected improvement method. Afterward, the results are verified by tension and compression tests. The results show that this method can substitute the complicated experimental tests which are normally employed to identify the extended Drucker-Prager parameters. Also, this method can be used to determine the mechanical properties of thermoplastic material such as tensile and compressive yield stresses and elastic modulus using only a three-point bending test. In addition, it is found that the volumetric change of thermoplastic during plastic deformation is significant and the non-associative, compared with the associative, plastic flow assumption is more proper for this material for the extended Drucker-Prager criterion.