Authors: Xinghui Zhao; Jiancong Fan; Yixuan Long
Addresses: College of Information Science and Engineering, Shandong University of Science and Technology, Qingdao, 266590, China ' College of Information Science and Engineering, Shandong University of Science and Technology, Qingdao, 266590, China ' University of Chinese Academy of Sciences, Beijing 100049, China; National Science Library, Chinese Academy of Science, Beijing 100190, China
Abstract: Based on the analysis of existing neighbourhood rough sets algorithm, a new attribute reduction algorithm, called Canopy-FCMVNRSMAR algorithm by reducing attribute using Canopy-FCM variable neighbourhood rough set model is proposed in this paper. The proposed algorithm is constructed by using attribute importance degree as the heuristic condition and makes the setting of neighbourhood value completely according to the distribution of data. So it avoids the disadvantages of setting the global neighbourhood value. The experimental results on open datasets on UCI show that the proposed algorithm can preserve fewer conditional attributes and improve the classification accuracy of data. What's more, it can extend the use of neighbourhood rough sets.
Keywords: neighbourhood rough set; unsymmetrical variable neighbourhood; attribute importance degree; global fixed neighbourhood.
International Journal of Collaborative Intelligence, 2019 Vol.2 No.1, pp.26 - 41
Received: 29 Dec 2016
Accepted: 01 May 2017
Published online: 14 Mar 2019 *