Identifying brown bear habitat by a combined GIS and machine learning method

摘要: In this paper we attempt to identify brown bear (Ursus arctos) habitat in south-western part of Slovenia, a country lying on the north-western-most edge continuous Dinaric-Eastern Alps population. The knowledge base (in form decision tree) for expert system identifying suitable habitat, was induced by automated machine learning from recorded sightings, and then linked GIS thematic layers subsequent habitat/non-habitat classification entire study area. accuracy tree classifier 87% (KHAT 73%). mostly agreed with existing domain knowledge. For area main factors considered be important were percentage forest (positive), proximity settlements (negative) elevation above see however did not account patch size. After filtering out patches smaller than 5000 ha GIS, increased 89% 77%). Whereas 88% within forests, only 33% all forests as habitat.