Semi-supervised attribute reduction for interval data based on misclassification cost

Abstract

Attribute reduction is a key issue in rough set theory which is widely used to handle uncertain knowledge. In reality, partially labeled interval data exist widely. So far, there are very few studies on partially labeled interval information systems. In this paper, we first define the concept of interval neighborhood by means of Kullback-Leibler (KL) divergence and standard deviation. Then a method is proposed to estimate the missing label by the nearest labeled objects to an unlabeled object and the cost of misclassification is constructed. Next a new entropy structure based on misclassification cost is proposed. After that, a semi-supervised attribute reduction method for partially labeled interval data is advanced. Finally, The rationality and validity of the method are demonstrated by experimental comparison.

TFD-IIS-CRMCB: Telecom Fraud Detection for Incomplete Information Systems Based on Correlated Relation and Maximal Consistent Block

“…Remark 3. In a complete information system (U, A), condition entropy H(D | B), mutual information H(D; B), and joint entropy H(D ∪ B) of attribute B and D are defined as [39] H…”