Searching for approximate description of decision classes
Lech Polkowski, Andrzej Skowron, Piotr Synak, Jakub Wróblewski · 2014
We discuss a searching method for synthesis of approximate description of decision classes in large data tables (decision tables). The method consists of the following stages: (i) searching for basic templates which are next used as elementary building blocks for decision classes description; (ii) performing templates grouping as a pre-processing for generalisation and contraction; (iii) generalisation and contraction operations performed on the results of grouping. The main goal of the method is to synthesize the approximate description of decision classes. The control in the searching method is focused on attempts to reduce uncertainty in approximate description of decision classes. On the other hand uncertainty in temporary synthesised descriptions of decision classes is the main driving force for the searching method. In the paper we concentrate on the different methods for template generation and on the discussion of performed computer experiments. We also present a general searching scheme for approximate description of decision classes as well as we point out the relationship of our approach to the rough mereological approach to the synthesis of complex objects. The presented method can be adopted for synthesis of adaptive decision algorithms what is the goal of our ongoing research project.