With the advances in technology and materials science, the use of composites has increased dramatically in various applications. Therefore, predicting the properties and behavior of composites under different conditions is of particular importance. Composites are mainly composed of four components: matrix, fibers, interphase and interface. The formation of these inter-phase zones has a great impact on the overall properties of the composite and due to the very small size and complexity of the behavior of this area and the uncertainty in existing experimental methods, the study of these materials is difficult. Therefore, the aim of the present study is to attempt to model this area and its effect on the final properties of polypropylene/glass fiber composite. For this purpose, due to limitations in computational models, stochastic analysis is used to predict the behavior of composite material. Initially, based on a study of the litrature on polypropylene/glass fiber composites and considering the probable distribution of effective parameters in the interphase and interface, the experimental design required for finite element modeling has been generated. Then, by producing a representative volume element in finite element software, micromechanical simulations are performed on the experimental design space. Then, to validate the micromechanical model, a polypropylene / glass fiber composite specimen was modeled and a good agreement was observed between the numerical results of this model and the existing results. Then, the results from the finite element software of the appropriate surrogate model are made and finally after validation of the surrogate model the sensitivity analysis is performed on the effectiveness of the responses to the individual inputs as well as with respect to the inputs. The interaction of their effects on each other is identified and finally the ineffective parameters can be eliminated to simplify the surrogate model. Keywords: Polypropylene/glass fiber composites, Interphase/Interface, Uncertainty, Surrogate model, Sensitivity analysis.