Improved evolutionary-based feature selection technique using extension of knowledge based on the rough approximations

This paper establishes an innovative approach of rough set (RS) approximations, namely the extension of knowledge based on the rough approximation (EKRA), which generalizes the old concepts and gets preferable results by reducing the boundary regions. In contrast to the former RS methods that obtained upper and lower approximations by several methods for special cases of binary relations. In addition, to assess the applicability of this approach it is combined with LSHADE with semi-parameter adaptation combined with CMA-ES (LSHADE-SPACMA) as a feature selection method, where EKRA is used as an objective function. The developed FS approach, named, LSPEKRA, which depends on LSHADE-SPACMA and EKRA aims to find the relevant features. This leads to improving the classification of different datasets. The experimental results show the great performance of the presented method against other Evolutionary algorithms. In addition, the FS methods based on EKRA provide results better than traditional RS in terms of performance measures. (C) 2022 Elsevier Inc. All rights reserved.

Automated summary

This paper introduces an innovative approach called Extension of Knowledge based on Rough Approximation (EKRA) which improves the results by reducing boundary regions. In the context of feature selection, EKRA is combined with LSHADE-SPACMA to form a method called LSPEKRA. Experimental results demonstrate the excellent performance of this method compared to other evolutionary algorithms and traditional RS methods.