Abstract
For decision tables whose condition and decision attributes are ordinal and monotonically interrelated, rule induction methods based on the dominance-based rough set approach has been proposed. The induced rules are used for class estimation. When the number of unclassified data is small and when the class evaluation changes dynamically, taking a lot of computational effort for rule induction is not always a good way. Therefore, without inducing rules, case-based reasoning approaches using rough membership values to decision tables in which dominance principle is assumed has been proposed. In this paper, we investigate the effects of parameter changes in k-nearest neighbor and the aggregations of rough membership values to the class estimation accuracy, and examine the effectiveness of the proposed case based reasoning methods.
