Dependency and accuracy measures for directed graphs

In this paper we use finite directed graphs (digraphs) as mathematical models to study two basic notions widely analyzed in granular computing: the attribute dependency and the approximation accuracy. To be more specific, at first we interpret any digraph as a Boolean information table, next we study the approximation accuracy for three fundamentals digraph families: the directed path, the directed cycle and the transitive tournament. We also introduce a new global average for the attribute dependency in any information table and we determine such number for any directed path. For the transitive tournament we provide a lower bound.