Research on Feature Selection Algorithms Based on Unbalanced Data

Abstract

The problem of unbalanced classification widely exists in medical,economic and other fields.Research shows that for the classification of unbalanced data sets,especially when the data is high-dimensional,an effective feature selection algorithm is crucial.However,most feature selection algorithms do not consider the impact of feature synergy,resulting in a decrease in classification performance.Considering the synergy of features,this paper proposes a new feature selection algorithm,FSBS,on the basis of the improved FAST feature selection algorithm.This algorithm employs AUC to evaluate the features,and the mutual gain is used to measure the magnitude of synergy.Then the effective features are selected and the unbalanced data are classified.Experimental results show that the proposed algorithm can effectively select features and improve the classification performance,especially when the number of features is small.