Genomic selection (GS) is a novel breeding strategy that selects individuals with high breeding value using computer programs. Although GS has long been practiced in the field of animal breeding, its application is still challenging in crops with high breeding efficiency, due to the limited training population size, the nature of genotype-environment interactions, and the complex interaction patterns between molecular markers. In this study, we developed a bioinformatics pipeline to perform machine learning (ML)-based classification for GS. We built a random forest-based ML classifier to produce an improved prediction performance, compared with four widely used GS prediction models on the maize GS dataset under study. We found that a reasonable ratio between positive and negative samples of training dataset is required in the ML-based GS classification system. Moreover, we recommended more careful selection of informative SNPs to build a ML-based GS model with high prediction performance.