Reducts Versus Constructs: an Experimental Evaluation

Abstract A reduct is one of the main notions in the Rough Sets Theory (RST). The idea of the reduct proved to be interesting enough to inspire a great deal of research resulting in various reduct-related notions and concepts. Reducts have been the subject of multiple papers, in which various approaches to the problem of defining, generating and applying reducts have been proposed. Given a data set, in which some pre-defined objects are described in terms of numerous parameters, the idea is to obtain a more compact description of these objects. As there are no universal descriptions for all possible applications, the idea is to focus on a description that is advantageous from a particular point of view, e.g. represented in the form of a consistency condition. Such a description may be created when sparse/redundant attributes are identified and eliminated from the data. In its classic form, reducts are minimal subsets of attributes that retain the consistency condition. The paper introduces the idea of constructs, in which the condition is modified so that the construct should manifest even better properties than those of reducts.