Method for selecting informative and non-informative features

Abstract

Abstract Today, solving the problems of choosing informative and non-informative features is an urgent problem, since the areas of application of methods and algorithms for pattern recognition to solve modern applied problems are expanding. But when solving these problems, problems arise that interfere with the effective use of well-known methods for choosing informative and non-informative features. The choice of a combination of a method and a criterion is one of these problems that is most suitable for the task, that is, with the search for such a pair of methods and criteria in which computational costs and recognition error are minimized. This article proposes a new method for choosing informative and non-informative features.