This paper describes a numerical optimization procedure for performance improvement of a tunnel ventilation jet fan. We employ optimization techniques based on a PRESS (predicted error sum of squares)-based-averaging (PBA) model to improve the aerodynamic performance of a tunnel ventilation jet fan. For numerical analyses, three-dimensional Reynolds-averaged Navier-Stokes (RANS) equations with a shear stress transport turbulence model are discretized using finite volume approximations and solved on hexahedral grids to evaluate the total efficiency at operating conditions as the objective function. Numerical results are verified by using experimental data for the total efficiency and the effective outlet velocity at operating conditions. Further, four geometric variables are used defining the meridional length and the thickness profile at the hub and shroud in the jet fan rotor as the design variables for numerical optimization. The results of numerical optimization show a significant improvement in the total efficiency of the optimum model in comparison with the base model.