Attribute Reduction Based on Bi-directional Distance Correlation and Radial Basis Network

  • Authors:

  • Li-Chao Chen;Wei Zhang;Ying-Jun Zhang;Bin Ye;Li-Hu Pan;Jing Li

  • Affiliations:

  • Institute of Computer Science and Technology, Taiyuan University of Science and Technology, No.66, Waliu road, Wanbolin District, Taiyuan Shanxi Prov., 030024, China;Institute of Computer Science and Technology, Taiyuan University of Science and Technology, No.66, Waliu road, Wanbolin District, Taiyuan Shanxi Prov., 030024, China;Institute of Computer Science and Technology, Taiyuan University of Science and Technology, No.66, Waliu road, Wanbolin District, Taiyuan Shanxi Prov., 030024, China;Institute of Computer Science and Technology, Taiyuan University of Science and Technology, No.66, Waliu road, Wanbolin District, Taiyuan Shanxi Prov., 030024, China;Institute of Computer Science and Technology, Taiyuan University of Science and Technology, No.66, Waliu road, Wanbolin District, Taiyuan Shanxi Prov., 030024, China;Institute of Computer Science and Technology, Taiyuan University of Science and Technology, No.66, Waliu road, Wanbolin District, Taiyuan Shanxi Prov., 030024, China

  • Venue:
  • ISNN '07 Proceedings of the 4th international symposium on Neural Networks: Part II--Advances in Neural Networks
  • Year:
  • 2007

Quantified Score

Hi-index 0.00

Visualization

Abstract

Attribute reduction is one of the important means to improve the efficiency and the quality of data mining, especially for high dimension data. From the view of distance and correction , the bi-directional distance and correction method was presented. This method can be used to measure the importance of dada attributes. Moreover, the revised decrease-increase combination strategy was used to reduce dimensionality and the radial basis neural network was used to validate the sub-set. This method adopts appropriate correlation function according to sample characteristic, which can avoid the limitation of IOC method. Since the longitudinal input-output connection and the horizontal difference between attribute and target was taken into account, the measure of the attribute importance will be more rational. So, quality data will be supply for the process of data mining subsequently.