The problem of convergence of classifiers construction procedure in the schemes of logical and algorithmic classification trees

Abstract

The paper considers the problem of convergence in the procedure of classifier schemes synthesis by methods of logical and algorithmic classification trees. It suggests the upper evaluation of complexity for the scheme of algorithms tree in the problem of approximation of real data array by a set of generalized features with a fixed criterion of termination of the branching procedure at the stage of the classification tree construction. This approach provides the required accuracy of the model, evaluates its complexity, reduces the number of branches, and achieves the required performance features. The paper represents the evaluation of convergence for the procedure of recognition schemes construction is logical and algorithmic classification trees on conditions of weak and robust separation of primary initial sampling classes.