Hybrid approaches in the context of data mining for predictive modeling in the bioeconomy and oil and gas industry

Aim. This work aimed to develop and validate a methodological approach to predictive modeling that, through the use of hybrid models (combining machine learning, statistical modeling, and expert systems), improves forecast accuracy and decision stability in the context of high data uncertainty and variability. Objectives. The work seeks to analyze Russian and international experience in applying data mining in the bioeconomy and oil and gas industries; to identify the limitations of individual methods (statistical models and neural networks) in forecasting complex nonlinear processes; to develop hybrid models for typical tasks, such as crop yield forecasting, assessing the potential for biotechnological waste processing, forecasting oil and gas production, and optimizing supply chains; and to evaluate their effectiveness compared to traditional approaches. Methods. The comparative study was conducted using open international databases (FAO, IEA, World Bank), corporate reporting (e.g., Shell, Gazprom Neft), and strategic documents. Hybrid models were applied to each case, namely statistical methods and machine learning algorithms were supplemented with expert rules. Data cleaning and reconstruction (including Bayesian missing data imputation), indicator normalization, and subsequent model training (gradient boosting, random forest, recurrent neural networks, etc.) were performed, incorporating expert knowledge at the model setup stage. Results. The proposed hybrid models demonstrated consistent advantages over traditional and single-model forecasting methods. There were improved accuracy and stability of results under conditions of limited, incomplete, and noisy data. Application of the developed approach to problems in the agro-industrial complex, biotechnological processing, and the oil and gas sector demonstrated its versatility and adaptability to various data types and production scenarios. The use of hybrid solutions improves forecasting efficiency, optimizes resource and logistics processes, and reduces operational risks. Positive economic and environmental effects are evident at the industry level, which are consistent with trends in international sustainable production and data management practices. Conclusions. Hybrid intelligent systems have demonstrated high efficiency and potential for strategic planning and operational management in the context of the global energy transition and biotechnological transformation. They provide more accurate and reliable forecasts, help reduce costs and risks, as well as improve the adaptability of managing the complex production and economic systems. The study results confirm the feasibility of widespread implementation of hybrid models in the bioeconomy and oil and gas industries, which will contribute to achieving goals of sustainable development and technological sovereignty.