A multistep methodology is explored for the optimization of the parameters of molecular models. This methodology is based on response surface methods and circumvents the practically impossible rigorous parameter fitting of computationally intensive molecular models. The approach is demonstrated for the parameters of the catalytic oxidation reaction of CO on platinum modeled by kinetic Monte Carlo simulations. After implementation of an initial reaction mechanism, sensitivity analysis is carried out to determine the key (active) reaction parameters for selected experimental points. Next, solution mapping is used to parametrize the model responses as low-degree polynomials of the active parameters. Finally, optimization of the active parameters is performed using simulated annealing. The optimized parameters are contrasted to those of a continuum-type mean-field model. The effect of surface diffusion in determining intrinsic kinetic parameters is also addressed, and the possibility of bridging the pressure gap via multiscale simulations is discussed.