The Improvement Of OFFSS Algorithm And Its Implementation

Feature subset selection is a difficult problem in pattern recognition and machine leaning, which aims to reduce the number of features used in classification or recognition. This reduction is helpful to improve the performance of classification algorithms in terms of speed, accuracy and simplicity. Most existing feature selection investigations focus on the case that the feature values are real or nominal, very little research is found to address the fuzzy-valued feature subset selection and its computational complexity. OFFSS (Optimal Fuzzy-Valued Feature Subset Selection) is an optimal feature subset selection based on fuzzy-valued extension matrix. Through computing the overlapping degree OV between the positive set and the negative set, threshold T is determined by the overlapping degree. OFFSS is a problem of searching for the feature subset S* which contains the least features in the condition of OV≤T. This paper gives an improved algorithm of OFFSS which adds expense of underT elements of extension matrix, and then applies Object Oriented software technology to implement OFFSS system which is based on GUI. Many experiments show the feasibility of the improved program.