Methods of Neural Network Modeling of Clusterization of Taxpayers to Determine Credit Risk by a Financial Regulator

Abstract

The subject of the research is devoted to the issues of building clustering models of enterprises - taxpayers in the tasks of tax administration and the financial sector. At the same time, the taxpayer clustering model is considered as a tool to support decision-making on credit risk by the financial regulator. The purpose of the author’s research is to develop correct management decisions in this task. Research methods require knowledge of a fairly reliable financial and economic condition of enterprises - taxpayers. The objective of the study is how, based on the set of economic indicators available to the tax authorities, on the basis of mathematical models, it is possible to carry out early diagnostics of unfavorable trends in the development of the financial condition of an enterprise, i.e., and possibly impending bankruptcy. The result of a computational experiment can be an array of tax declarations for a certain preceding time period for a set of taxation objects of interest to the analyst using neural network modeling. On the basis of such a neural network model, enterprises - taxpayers can be divided into a certain number of clusters, in terms of the task. When developing management decisions, it will be more effective in the aspect of simultaneously taking into account the interests of the budget, preserving the economic stability of enterprises - taxpayers and the financial regulator, if, as a result of the experiment, the regulator receives sufficiently reliable information about the belonging of a particular enterprise to one of the clusters proposed in the problem. In the current post-image situation, when survival reserves and credit reserves are interconnected, neural network models in artificial intelligence show a fairly accurate analysis result.