Forecasting the risks of non-implementation of national project activities

Abstract

The article addresses the problem of monitoring and forecasting risks of non‑fulfilment of activities within national projects (NP) and federal projects (FP) in the framework of the State Automated Information System «Management» (GASU). The aim of the study is to create a technological solution for the formation of tools for modeling and forecasting the failure of national and federal projects based on machine learning methods, providing a high-quality forecast for federal decision-making systems. A structurally stable risk forecasting model for delays in milestone achievement (MA) has been developed. The model is based on an ensemble of machine learning methods and leverages key indicators and data on milestones, including financial metrics and physical (natural) performance indicators. The results of the classification using an ensemble of machine learning methods were integrated with data on the financial risks of events and delays admitted earlier in the year under review. The proposed model was validated using real‑world data for 2023–2024 and subsequently deployed into industrial operation within the GASU in 2025. Experimental results demonstrate superior performance in terms of quality metrics when forecasting at the activity level, as compared to the level of individual milestones. A promising direction for further development involves expanding the feature space through additional analytics — incorporating budgetary, tax‑related, methodological, and other types of risks — as well as enhancing forecasting capabilities to estimate the level of activity accomplishment. The developed technological solution will improve the quality of decision-making that is of significant importance in the area of managing national and federal projects.