An Inductive Methodology for Data-Based Rules Building

Abstract

Extraction of rules from databases for classification and decision tasks is an issue of growing importance as automated processes based on data are being required in these fields. An inductive methodology for data-based rules building and automated learning is presented in this paper. A fuzzy framework is used for knowledge representation and, through the introduction and the use of dual properties in the valuation space of response variables, reasons for and against the rules are evaluated from data. This make possible to use continuous DDT logic, which provides a more general and informative framework, in order to assess the validity of rules and build an appropriate knowledge base.