TUNING OF REAL TYPE DATA CODING FOR DECISION TABLES

摘要

The rough set theory can be applied to knowledge acquisition from imprecise data. The input data are presented in a decision table, where rows correspond to objects and columns are related to attributes describing features of objects. The attributes are divided into conditional ones and decisions. The knowledge is represented in a form of decision rules, denoting relations between conditional attributes and decisions. Some values of attributes can be numbers of real type. The values are submitted to process of discretiazation and then they are encoded. The paper presents an algorithm of optimization of real type data discretization. The arbitrary scheme of discretiazation (proposed by an expert) is tuned to improve the quality of rules derived from the decision table.