Background: In recent years, a growing body of literature has examined the impact of neighborhood characteristics on Alzheimer's disease (AD) dementia. However, spatial variability of the most influential variables and their relative importance to AD dementia prevalence remain underexplored. Methods: We compiled various widely recognized factors to examine spatial heterogeneity and associations with AD dementia prevalence utilizing non-linear geographically weighted random forest approach. Results: The model outperformed conventional ones, with an out-of-bag R2 of 74.8%. Key findings showed the normalized difference vegetation index as the most influential environmental factor in 15.1% of US counties, lack of leisure time physical activity in 12.7%, binge drinking in 9.1%, and mobile homes as the main socioeconomic factor in 13.3% of US counties. Discussion: Spatial machine learning analyses suggest that AD dementia prevalence could be impacted by county-specific targeted interventions that improve green space, reduce air pollution and support greater access to physical activity.