URBAN ECOLOGICAL MAPS must be created by local governments by NATURAL ENVIRONMENT CONSERVATION ACT, and the maps are generally called biotope map. So far, biotope maps study was a tendency to focus on the type of vegetation, naturalness, land use, landscape ecology theories. However, biotope related studies have not reflected the concept of animal habitat, which is a component of biotope, and that is the limitation of biotope map research.
This study suggest a methodology to predict potential habitats forfauna using machine learning to quantify habitat values. The potential habitats of fauna were predicted by spatial statistics using machine learning, and the results were converted into species richness. For biotope type assessments, we classified biotope values into vegetation value and habitat value and evaluated them using a matrix for value summation. The vegetation value was divided into 5 stages based on vegetation nature and land use, and the habitat value was classified into five stages by predicting the species richness predicted by machine learning. This is meaningful because our research can positively reflect the results of field surveys of fauna that were negatively reflected in the evaluation of biotope types in the past. Therefore, in the future, if the biotope map manual is revised, our methodology should be applied.