Supervised Feature Subset Selection using Extended Fuzzy Absolute Information Measure for Different Classifiers

Feature subset selection plays an important role in data mining and machine learning applications. The main aim of feature subset selection is reducing dimensionality by removing irrelevant and redundant features and improving classification accuracy. This paper presents a supervised feature selection method called as Extended Fuzzy Absolute Information Measure (EFAIM) for different classifiers. In this process, first discretization algorithm is applied to discretize numeric and nominal features of a database to construct fuzzy sets of a feature. Then the method EFAIM is applied to select feature subset focusing on boundary samples. To verify the effectiveness of this method, several experiments are conducted for different classifiers, such as, LMT, Naïve Bayes, SMO, C4.5, JRIP, PART and Simple Cart with different UCI datasets. The Experimental results indicates that the proposed algorithm have achieved better classification accuracy for all datasets, that is, almost above 75% of accuracy. For WINE dataset, it gets 96% of classification accuracy for Naïve Bayes classifier. For Ionosphere dataset, it gives almost 89% of classification accuracy for maximum of classifiers with minimum selected feature subset. Thus improved classification accuracy is obtained with selected subset of minimum number of features at minimum processing time.