Article ID: | iaor20022070 |
Country: | Netherlands |
Volume: | 134 |
Issue: | 3 |
Start Page Number: | 592 |
End Page Number: | 605 |
Publication Date: | Nov 2001 |
Journal: | European Journal of Operational Research |
Authors: | Beynon Malcolm |
Keywords: | statistics: empirical, datamining |
One fundamental aspect of the variable precision rough sets (VPRS) model involves a search for subsets of condition attributes which provide the same information for classification purposes as the full set of available attributes. Such subsets are labelled ‘approximate reducts’ or ‘β-reducts’, being defined for a specified classification error denoted by β. This paper undertakes a further investigation of the criteria for a β-reduct within VPRS. Certain anomalies and interesting implications are identified. An additional condition is suggested for finding β-reducts which assures a more general level knowledge equivalent to that of the full set of attributes.