Feature subset selection is one of the widely used and practical methods for classification and pattern recognition, which aims to reduce the computational complexity, improve the speed of computation, classification accuracy in classification and recognition problem by reducing the dimensionlity of features. Optimal fuzzy-valued feature subset selection (OFFSS) method has been efficient for feature subset selection of two-class problem, which can be classified into positive and negative classes. However, it is not suitable for multi-class problem, i.e. Multi-class Optimal Fuzzy-valued Feature Subset Selection (MOFFSS). In this paper, OFFSS algorithm is extended to for multi-class problem by two techniques where the information entropy is used to reduce computational complexity. The feasibility and simplicity of the two improved algorithms are demonstrated by applying the feature subset selected to fuzzy decision tree induction.
|