Habitat use by animals (foraging, resting and so on) is the result of habitat selection. It is dependent on the spatial scales. Crop damage by wildlife which is serious problem in Japan is also the result of animal habitat selection. To resolve this problem, wildlife management in consideration of the concept of spatial scale is necessary. The objective of this study was to evaluate the difference of environmental factors which contributed to crop-raiding Japanese macaques (Macaca fuscata) using multi-scale analysis. We used the geo-information of crop damaged and non-damaged area for predicting the risk of crop damage. Environmental factors related to crop damage were defined as 16 variables: Ratio of broad-leaved forests, conifers, grass, and urban, the minimum distance from forest edge, broad-leaved, conifers, grass, urban, water area, road, fence, and alarm, the shape of forest edge, the number of fence and alarm. We defined multi buffers (100m-2500m) as spatial extent. Random Forest classification was applied for predicting the risk of crop damage. Using 10-hold cross validation, the accuracies of model were evaluated. From the results, the importance of variable was changed following the spatial extent. The scale dependence of crop damage by Japanese macaques was suggested from the multi scale analysis. Finally, we will propose the effective countermeasures in each spatial scale.