Classification of objects in grid computing

Abstract

This study reviews the problem of creating an algorithm for classifying objects in a Grid system, which will allow the automation of this process to proceed. Splitting objects into classes allows you to effectively manage the resources and processes that occur in a distributed system. Classification according to the current values of characteristics, by reference values of characteristics, with an analysis of the original causes of the result and also a modeled classification is considered. The problem of determining non-informative features is being solved. The problems of determining the function of the dependence of a multidimensional feature space for solving classification problems are considered. The concept of a reference state of a feature is introduced in which it is not amenable to the influence of other features. A procedure is proposed for determining the function of the dependence of characteristics having a reference state. Using the method of standardized multiple regression equations, the optimal variant of the factors included in the model was chosen, since the factors with the lowest value have the least influence on the dependent parameter and can be excluded.