Accurate prediction of the vibration response at a point on a complex structure, where the operational behavior cannot be measured directly, is an important engineering problem for design optimization, component selection and condition monitoring. Identifying the exciting forces acting on the structure is a major step in the vibration response prediction (VRP). At the point where direct measurement is impossible or impractical due to physical constraints, a common approach is to identify the exciting forces based on multiplication of an inverted frequency response function (FRF) matrix and a vector of vibration responses measured at the points where the exciting forces are transmitted through. However, in some cases measuring FRFs are almost impossible. In other cases, where measuring is possible, they may be prone to significant errors. Furthermore, the inverted FRF matrix may be ill-conditioned due to the one or few modes that dominate the dynamics of the structure.In order to improve the force identification step and reduce the experimental challenges, previous studies focused on either conditioning methods or numerical models. However, conditioning methods result in additional measurements, and using only numerical models causes reduced accuracy due to incongruities between the simulation model and the real system. Considering these problems, a hybrid VRP methodology that incorporates the numerical modeling and experimental measurement results is proposed in this study. Creating an accurate numerical model and properly selecting the force identification points are the main requirements of the proposed methodology. A structure coupled with rubber mounts is used to demonstrate the proposed methodology. The numerical model includes hyperelastic and viscoelastic modeling of the rubber to represent the system behavior accurately. The selection of force identification points is based on a metric that is composed of the average condition number of the FRF matrix across the whole frequency of interest. The results show that the proposed hybrid methodology is superior to other alternative methods where predictions are solely based on numerical results or experimental measurements.