To solve the problem of material parameters identification in the non-linear range, an inverse method based on an optimization technique and a finite element method is proposed. Associated with the finite element calculation, a sensitivity analaysis is developed on the base of a direct differentiation method. The frame of this analysis is consistent with the algorithm used in stress calculation which is based on an hyperelastic formulation of the constitutive equations and a prediction-correction scheme with a multiplicative split of the deformation gradient. Application of this inverse identification algorithm to the determination of viscoplastic material parameters of an aluminium alloy is presented. In this application, the effect of experimental errors, the effect of the initial guess as well as the effect of the number of experimental tests used is studied and used to determine confidence intervals on the identified parameters.