Reproducing dynamical properties, such as diffusion coefficients, in coarse-grained (CG) molecular dynamics simulations can be challenging due to the loss of fine-grained details, such as atomic vibrations and local motions of particles in the parent all-atom (AA) system. In this study, we present a predictive tool for the mobility acceleration factor, defined as ratio of the CG diffusion coefficient to the AA diffusion coefficient. According to the well-established Green-Kubo formalism, the diffusion coefficient is related to integral of the velocity autocorrelation function. As integral of the velocity autocorrelation function is influenced by the particle's acceleration, key parameters affecting the acceleration differences between an AA molecule and its corresponding CG bead are identified to develop a predictive model. By conducting AA and CG simulations on 20 liquid hydrocarbons with varying masses and sizes, their mobility acceleration factors are determined, the largest being 62.78. This data is then used to fit a nonlinear functional form as the predictive model. The identified molecular descriptors for the predictive model are easy to calculate for new molecules, enabling the model to be readily applied to predict the mobility acceleration factor for different molecules in CG simulations.