Class-specific information measures and attribute reducts for hierarchy and systematicness

Attribute reduction of rough set theory underlies knowledge acquisition and has two hierarchical types (classification-based class-specific attribute reducts) perspectives from algebra information theory; thus, there are four combined modes in total. Informational reducts fundamental but lacking thus investigated by correspondingly constructing measures. First, three measures (i.e., entropy, conditional mutual information) novelly established at the class level decomposition to acquire their connection, systematical relationship, uncertainty semantics, granulation monotonicity. Second, informational proposed internal basic properties, heuristic algorithm. Third, achieve transverse connections, including strength feature consistency degeneration, with algebraic balance, classification-based reducts. Finally, relevant effectively verified decision tables data experiments. Class-specific deepen existing a isomorphism, while systematically perfect viewpoint isomorphisms; these results facilitate measurement processing, especially level.