A Neighborhood Rough Classification with Dempster-Shafer Evidence Theory

Abstract

In order to further improve the classification mechanism and the performance of neighborhood classifier, a Dempster-Shafer(D-S) evidence-driven neighborhood rough classification method is proposed. Firstly, in attribute reduction, the error rate of neighborhood decision is used as the index of attribute significance, and the attribute reduction method based on neighborhood decision error rate is studied. By removing redundant attributes, an important set of attributes is provided for classification learning. Then, in terms of classifier design, the traditional majority voting mechanism is revised, D-S evidence theory is introduced into the information fusion of neighborhood samples,and a neighborhood classifier based on D-S evidence theory is proposed. Experimental results on UCI public data set show that the proposed method has higher classification accuracy than the neighborhood classifier under the majority voting mechanism. The paper provides a new insight for the further study of neighborhood classification methods.

Keywords

• neighborhood decision error rate
• dempster-shafer evidence theory
• attribute reduction
• rough sets
• classification