Relational methodology for data mining and knowledge discovery

摘要

This paper analyses capabilities of machine learning and KDD&DM methods to perform cognitive processes in the form of discovering the domain theories. The concept of cognition of the domain theory is derived for the reprehensive measurement theory (RMT). We show that a relational data mining approach we proposed previously performs cognition of domain theories in accordance with the RMT and produces the relational methodology for analysis of cognitive capabilities of data mining methods. In this methodology a domain theory includes a metadata ontology. This ontology contains various data types formalized in the first-order logic in accordance with the RMT. To represent the knowledge theory we use the concept of the logical empirical theory that is defined in the paper.