Genomic prediction has become an increasingly popular tool for hybrid performance evaluation in plant breeding, mainly because that it can reduce cost and accelerate a breeding program. In this study, we propose a systematic procedure to predict hybrid performance using a genomic selection (GS) model that takes both additive and dominance marker effects into account. We first demonstrate the advantage of the additive-dominance effects model over the only additive effects model through a simulation study. Based on the additive-dominance model, we predict genomic estimated breeding values (GEBVs) for individual hybrid combinations and their parental lines. The GEBV based specific combining ability (SCA) for each hybrid, and general combining ability (GCAs) for its parental lines are then derived to quantify the degree of mid-parent heterosis (MPH) or better parent heterosis (BPH) of the hybrid. Finally, we estimate the variance components due to additive and dominance gene action effects, and heritability using a genomic BLUP model. These estimates are used to justify the results of the genomic prediction study. A pumpkin data set is given to illustrate the provided procedure. The data set consists of 320 parental lines with collected 61,179 SNP markers; 119, 120, 120 phenotypic values of hybrids on three quantitative traits within C. maxima; and 89, 111, 90 phenotypic values of hybrids on the same three quantitative traits within C. moshata.