USE OF MATHEMATICAL APPARATUS IN PUBLIC ADMINISTRATION OF FINANCIAL AND ECONOMIC PROCESSES AT THE NATIONAL LEVEL

Abstract

The work is devoted to the formation of a scientific basis for the justification of the use of mathematical apparatus in the public management of financial and economic processes at the national level. The work proves that most processes at the macro level of the economy have a cyclical nature of development. Forecasting the beginning and end of the cycle is an urgent task today, as it provides an opportunity to make reasonable and objective decisions at all levels of public administration. It is this task of identifying cycles in the macroeconomic system that is considered in the paper. In the study, the task was set, based on statistical data on the flow of real financial and economic processes at the national level, it is necessary to build a probabilistic model for identifying cycles based on the principle of data classification and establish the effectiveness of using different classification strategies and investigate the quality of classification by comparing the obtained results with expert assessments. The article proposes the concept of using Bayesian networks to identify (recognize) phases of the development of processes of an arbitrary nature, represented by time series of data, in particular in economics and finance. Previous studies carried out in this direction indicate the possibility of obtaining acceptable positive results. For this purpose, it is necessary to create special probabilistic models and estimate their parameters using statistical data regarding the development of the process. In order to analyze the presence of cycles in processes at the macro level, it is proposed to use a modern probabilistic modeling apparatus – Bayesian networks. Based on statistical data, a network structure is built, which is used to detect cycles at a given time interval. The correctness of the obtained result is confirmed by expert assessments. To determine availability of possible cycles in financial and economic systems is it proposed to use dynamic Bayesian network that represent an effective instrument for analysis of data of arbitrary nature. As a result of the network training and its use for the cycles identification it was established that the classification errors does not exceed 16.5% what is quite acceptable in this case. The directions of scientific research and further research regarding the substantiation of the possibilities of using Bayesian networks for modeling and forecasting the development of public management processes at various levels are outlined. In particular, establishing the possibility of their application to solving the problems of managing resources and risks associated with the implementation of management decisions.