Extending extension theory for classifying data with numerical values

作者：Liu Cheng Hsiang^*

来源：Neural Computing & Applications, 2013, 23(1): 161-167.

DOI：10.1007/s00521-011-0795-z

• Extension theory
• Classical domain
• Classification
• Numerical data set

The extension theory (ET) is one of the simplest and most attractive pattern classification methods. However, it has difficulty determining the classical domain. In addition, the traditional extended relational function used in extension theory does not provide very useful summaries of asymmetrical data. This study proposes a modified extension theory (MET) to overcome these shortcomings. The MET applies the largest sphere concept to determine the range of the classical domains and incorporates the information about the data distribution when calculating the relevance of an element belonging to a set. Experimental results indicate that the MET consistently achieved better or comparable results than the traditional ET. The MET also produces a classifier with satisfactory classification accuracy compared with well-known classifiers (e.g., decision trees and k-nearest neighbor).

• 出版日期2013-7