Dealloyed gold nanoparticles can be synthesized by selectively dissolving Ag from gold-silver alloy nanoparticles through the well-known process of dealloying. These nanoparticles exhibit remarkable catalytic activity towards the CO oxidation reaction owing to their large specific surface area, presence of rough surfaces that contain a high density of catalytically-active sites, and synergistic effects arising from the residual Ag leftover from the dealloying process. We shall introduce a computational framework aimed at determining the optimal dealloying conditions that can yield dealloyed nanoparticles with the highest CO oxidation specific activity. This will be accomplished by establishing using a comprehensive machine learning synthesis-structure-activity relations for dealloyed nanoparticles. Through such an approach, we elucidate the link between the key catalyst attributes (active site density, reaction kinetics, and synergistic effects) and dealloying parameters (initial nanoparticle size, composition, dissolution condition, and dissolution time). Our methodology combines multiscale simulations to explore a vast synthesis-parameter space. The resulting size-dependent synthesis-structure-activity map reveals the optimal residual Ag for efficient CO oxidation catalysis. If possible experiments will also be performed.