Authors: Changzhong Wang, Mingwen Shao, Qiang He, Yuhua Qian, Yali Qi

Abstract:

Fuzzy rough set is an important rough set model used for feature selection. It uses the fuzzy rough dependency as a  criterion for feature selection. However, this model can merely  maintain a maximal dependency function. It does not fit a given  data set well and cannot ideally describe the differences in sample  classification. Therefore, in this study, we introduce a new model  for handling this problem. First, we define the fuzzy decision of a  sample using the concept of fuzzy neighborhood. Then, a parameterized fuzzy relation is introduced to characterize the fuzzy  information granules, using which the fuzzy lower and upper  approximations of a decision are reconstructed and a new fuzzy  rough set model is introduced. This can guarantee that the  membership degree of a sample to its own category reaches the  maximal value. Furthermore, this approach can fit a given data  set and effectively prevents samples from being misclassified.  Finally, we define the significance measure of a candidate attribute and design a greedy forward algorithm for feature selection.  Twelve data sets selected from public data sources are used to  compare the proposed algorithm with certain existing algorithms,  and the experimental results show that the proposed reduction  algorithm is more effective than classical fuzzy rough sets,  especially for those data sets for which different categories exhibit  a large degree of overlap

Keywords： Dependency function; Fuzzy rough set; Fuzzy similarity relation; Feature selection

Feature subset selection based on fuzzy neighborhood rough sets.pdf

查看详情

Fri Aug 26 14:50:00 CST 2016