Good Classification Tests as Formal Concepts

摘要

The interconnection between the Diagnostic (Classification) Test Approach to Data Analysis and the Formal Concept Analysis (FCA) is considered. The definition of a good classification test is given via Galois's correspondences. Next we discuss the relations between good tests and formal concepts. A good classification test is understood as a good approximation of a given classification on a given set of examples. Classification tests serve as a basis for inferring implicative, functional dependencies and association rules from datasets. This approach gives the possibility to directly control the data analysis process by giving object classifications.