Classifiers based on approximate reasoning schemes

We discuss classifiers [3] for complex concepts constructed from data sets and domain knowledge using approximate reasoning schemes (AR schemes). The approach is based on granular computing methods developed using rough set and rough mereological approaches [9, 13, 7]. In experiments we use a road simulator (see [15]) making it possible to collect data, e.g., on vehicle-agents movement on the road, at the crossroads, and data from different sensor-agents. We compare the quality of two classifiers: the standard rough set classifier based on the set of minimal decision rules and the classifier based on AR schemes.