FROM DATA TO DECISIONS: THE ARCHITECTURE OF A NEURAL NETWORK MODEL FOR FORECASTING LABOR PROCESSES IN MEGACITIES

Relevance of the study. Digitalization is rapidly transforming social and labor relations in megacities, increasing the need for new approaches to their management. Traditional analysis tools cannot cope with the increasing complexity and volumes of data, while neural network technologies are capable of providing adaptability and accuracy in forecasting and regulating labor processes. The lack of uniform methods for preparing and structuring data significantly reduces the efficiency of implementing AI solutions in human resource management, which makes this study highly relevant. The aim of the work is to develop a methodological approach to defining and preparing parameters for creating a neural network model for managing social and labor relations in the context of digitalization of megacities. The main attention is paid to the formation of a system of relevant indicators, data normalization and integration of digital maturity parameters. Data and methods. The study used data from the World Bank, an analytical report on the digital maturity of economic sectors of the Russian Federation, as well as indicators of state statistics of megacities. The methods of correlation analysis, data normalization, factor selection of features, expert assessment, as well as formalized criteria based on weighting coefficients were used. For a preliminary assessment of the quality of the parameters, testing was used on pilot samples. Results. 10 key categories of parameters were identified, including digital literacy, urbanization level, employment, access to AI infrastructure and strategic planning in the social sphere. An algorithm for step-by-step data preparation for a neural network model has been developed, including defining goals, collecting and filtering parameters, normalization, analyzing relationships, and testing on pilot samples. It has been established that the optimal volume of parameters for effective model training is 12–15 indicators. The importance of integrating technological and socio-economic factors to improve forecast accuracy has been shown. Conclusions. The study confirmed the need for multi-level and interdisciplinary data preparation for building neural network models in the field of managing social and labor relations in megacities. The developed methodology allows for increasing the accuracy of modeling, adapting management to the rapidly changing conditions of the digital economy, and forms the basis for creating intelligent decision support systems in social policy. The results obtained have practical significance for government bodies and can be used in developing digital platforms for labor resource management.