Feature selection based on fuzzy extension matrix for multi-class problem

Feature subset selection is one of the widely used and practical methods for pattern recognition and classification, which aims to reduce the number of features to be used. Optimal fuzzy-valued feature subset selection (OFFSS) method is efficient for feature subset selection of two-class problem. However, the original OFFSS is not suitable for multi-class problem. This paper gives an improved version of OFFSS. The OFFSS algorithm is extended to the multi-class problem in which information entropy is used to reduce computational complexity of the method. The feasibility and simplicity of the improved algorithm are demonstrated by applying it to fuzzy decision tree induction.