Abstract
In this paper, we apply ensemble learning approaches to building a classifier with decision rules induced by a rough set based algorithm. More concretely, ensemble learning such as Bagging, Random Forest and Attribute Sampling Ensemble are utilized to build a classifier with decision rules induced by MLEM2 algorithm. We demonstrate the effectiveness of these ensemble learning methods in improvement of classification accuracy through numerical experiments. For utilizing the ensemble learning methods, the sample sizes should be determined. A method for determining the attribute sample size is proposed and examined its usefulness. Furthermore, the robustness of the proposed classifiers against missing attribute values is shown.
