Reviewing RELIEF and its extensions: a new approach for estimating attributes considering high-correlated features

Abstract

RELIEF algorithm and its extensions are some of the most known filter methods for estimating the quality of attributes in classification problems dealing with both dependent and independent features. These methods attend to find all meaningful features for each problem (both weakly and strongly ones) so they are usually employed like a first stage for detecting irrelevant attributes. Nevertheless, in this paper we checked that RELIEF-family algorithms present some important limitations that could distort the selection of the final features' subset, specially in the presence of high-correlated attributes. To overcome these difficulties, a new approach has been developed (WACSA algorithm), which performance and validity are verified on wellknown data sets.