The sample and instance selection for data dimensionality reduction

Abstract

The paper proposes tools for data dimensionality reduction containing sample selection method and instance informativity indicators based on the evolutionary search, which is modified to speed up the search through the creation of special operators, taking into account a priori information about the data sample and concentrating search on the most perspective solution areas. This allows preserving the stochastic nature of the search to accelerate the obtainment of acceptable solutions through the introduction of deterministic component in the search strategy. The proposed methods are experimentally studied. On the results of experiments the comparative characteristics and recommendations for the use of the proposed methods are given.