Geographic heterogeneous of both species-environment and inter-specific relationships play important roles in shaping the ecological characteristics of wildlife habitat selection and population distribution. However, how geographic heterogeneous affect the distribution of a target species in predator-prey systems, especially in human-dominated landscapes, is still unclear. Based on line transect survey and monitoring network, this study examined the spatial heterogeneity of habitat selection of Amur tiger and leopard, as large carnivores, and their main ungulate prey, wild boar and roe deer, in Northeast China, by applying logistic Geographically Weighted Regression (GWR) in conjunction with global logistic regression (GLM). Our results suggested that, the geographic heterogeneity effects of habitat factors on the habitat selection of target species are manifested in the intensity and trends of the heterogeneous effects. Furthermore, ungulate prey exhibit preferences for particular types of habitats, subsequently promoting the distribution patterns of large felids. Our findings highlight the significance of spatial non-stationarity in ungulate prey habitat selections, and this may be the key to promoting large carnivores adapting to habitats around roads and reducing habitat isolation and reduction caused by road avoidance of ungulates.