Silver Chub (Macrhybopsis storeriana) was once a common forage fish in Lake Erie but has declined greatly since the 1950s. Identification of optimal and marginal habitats would help conserve and manage this species. We developed neural networks to use broad-scale habitat variables to predict abundance classes of Silver Chub in western Lake Erie, where its largest remaining population exists. Model performance was good, particularly for predicting locations of habitat with the potential to support the highest and lowest abundances of this species. Highest abundances are expected in waters >5 m deep; water depth and distance to coastal habitats were important model features. These models provide initial tools to help conserve this species, but their resolution can be improved with additional data and consideration of other ecological factors.