New Developments of Determinacy Analysis

This paper deals with development of Determincy Analysis (DA), a method for knowledge discovery. There are three approaches to DA that give different results. The first method finds exactly one system of non-overlapping rules, all with equal length. The second, step by step approach enables to find very many different systems of non-overlapping rules where the rules have different length. The third approach can find one system of overlapping rules. Analysis of these approaches showed that shorter rule systems contain rules which are not contained in other rules. It means that we have to find only all these rules (so called determinative set of rules, DSR) and describe given data table on their basis. This paper presents a new DA approach based on the DSR and algorithm for finding them. Several new tasks are formulated: to generate rule systems with specific features or the shortest or minimal rule system etc.