Automatic classification of location contexts with decision trees

Abstract

Location contexts are geographic regions, with well defined boundaries, that can be used to characterize the context of the persons lying inside them. In this paper we describe a process that exploits the increasing availability of geographic data to automatically create and classify location contexts. The pro-posed process generates new geographic regions from a database of Points Of Interest through the use of spatial clustering techniques, and classifies them automatically using a decision tree based method. Some preliminary results demonstrate the validity of this approach, while suggesting that a richer geographic database could produce location contexts of higher quality.