The ways of introducing AI/ML-based prediction methods for the improvement of the system of government socio-economic administration in Ukraine

The objective of the article is to develop and test in practice a mechanism for constructing AI/ML-based predictions, adapted for use in the system of government socio-economic administration in Ukraine. Research design is represented by several methods like qualitative analysis in order to identify potential benefits of AI use in different spheres of government administration, synthesis to generate new datasets for the experiment, and abstraction to abstract from the current situation in Ukraine, population displacement, uneven statistics reporting. Among empirical methods are prediction and experimental methods to construct a mechanism for the implementation of AI/ML prediction methods in public administration, develop a high-level architecture of the AI/ML prediction system, and create and train the COVID-19 prediction neuron network. A holistic vision of the AI/ML-based prediction construction mechanism, depending on data taken from state official online platforms, is presented, in addition, the ways of its possible practical application for the improvement of the national system of state socio-economic administration are described. The main condition and guarantee of obtaining accurate results is access to quality data through platforms such as Diia, HELSI, national education platforms, government banks, etc. The findings of the research suggest that wide implementation of AI/ML-based prediction technologies will allow the government in perspective to increase the efficiency of the use of budgetary resources, the effectiveness of the government target programs, improve the quality of public administration and to better satisfy the citizens’ demand. Future studies should be done to overcome the limitations of the approach: find a way to protect and extract sensitive information from government platforms, fight neural network bias, and create a more perfect system that is able to make multiparameter predictions and is also self-improving on the basis of the obtained results.