Constructing associative classifier using rough sets and evidence theory

Abstract

Constructing accurate classifier based on association rule is an important and challenging task in data mining. In this paper, a novel combination strategy based on rough sets (RST) and evidence theory (DST) for associative classification (RSETAC) is proposed. In RSETAC, rules are regarded as classification experts, after the calculation of the basic probability assignments (bpa) according to rule confidences and evidence weights employing RST, Yang's rule of combination is employed to combine the distinct evidences to realize an aggregate classification. A numerical example is shown to highlight the procedure of the proposed method. The comparison with popular methods like CBA, C4.5, RIPPER and MCAR indicates that RSETAC is a competitive method for classification based on association rule. © Springer-Verlag Berlin Heidelberg 2007.