Authors : Elsayed Radwan and Y. Alkazemi Basem
Abstract: Knowledge extraction from various datasets is suffering from missing attribute values, representing the non-categorical ones, besides the inconsistency among the local rules. These shorts reflect two main problems, information granules representation and superfluous feature existence. This study treats these problems based on tolerance rough sets. A similarity relation is considered based on the matching percentage among objects in order to decrease the knowledge granularity and hence, increase the information capacity. In order to measure the dependencies among features, a tolerance measure based on step-deck coordination function is defined as vague inclusion function. For feature extraction process, a consistency function is defined. It is proved that the consistency measure promise an increase in classification accuracy. More specifically, the proposed model receives a base method of tolerance rough set to generate local rules that describe the application algorithm. The details and limitations of the proposed model are discussed where a comparative study is appeared.
Elsayed Radwan and Y. Alkazemi Basem, 2018. Extracting Knowledge from Incomplete Dataset by Tolerance Rough Sets with Consistency Measure. International Journal of Soft Computing, 13: 149-157.