Contents I. Introduction
An integrated rating mechanisms (hereinafter referred as IRM) [1], [2] consisting of a set of criteria scoring scales Xi, a criteria tree with binary structure describing by graph G ∈ G, a set of convolution matrices M and discrete integration procedure P, can be identified based on training set. Particularly, in [3], [4] identification problem of IRM was investigated as optimization problem. That approach based on the representation of the IRM in the form of a superposition of the products of one-hot encoded criteria and matrices. In [5] was presented idea of sensitivity of training set to subset of input criteria. This approach was used for searching equivalency sets of particular criteria and selection of tree structure.
