¹School of Computer Science and Technology, China University of Mining and Technology, Xuzhou, China 221116
²School of Medical Information, Xuzhou Medical College, Xuzhou, China 221000
³Key Laboratory of Intelligent Information Processing, Institute of Computing Technology, Chinese Academy of Sciences, Beijing, China 100190

---

Abstract—For high dimensional data, the redundant attributes of samplers will not only increase the complexity of the calculation, but also affect the accuracy of final result. The existing attribute reduction methods are encountering bottleneck problem of timeliness and spatiality. In order to looking for a relatively coarse attributes granularity of problem solving, this paper proposes an efficient attribute granulation method to remove redundancy attribute. The method calculates the similarity of attributes according attribute discernibility first, and then clusters attributes into several group through affinity propagation clustering algorithm. At last, representative attributes are produced through some algorithms to form a coarser attribute granularity. Experimental results show that the attribute granulation method based on affinity propagation clustering algorithm(AGAP) method is a more efficient algorithm than traditional attribute reduction algorithm(AR).

Index Terms—Attribute granulation, attribute dependability, AP clustering, parallel computing.

[PDF]

Cite: Hong Zhu, Shifei Ding, Han Zhao, Lina Bao, "Attribute Granulation Based on Attribute Discernibility and AP Algorithm," Journal of Software vol. 8, no. 4, pp. 834-841, 2013.