Neural network forecasting of the agrifood system performance: Theoretical and methodological foundations

Agrifood systems are increasingly subject to economic instability caused by a wide range of factors, which are difficult to identify and thus impede the forecasting of the systems’ functioning. The purpose of the article is to form theoretical and methodological foundations of the neural network forecasting of agrifood systems. Methodologically, the research rests on the integration theory as applied to enterprises belonging to agroindustrial complex, and the systems approach. A range of general scientific and special methods are used, including elements of statistical analysis. The paper provides an original definition to the term “agrifood system” interpreting it as a set of interconnected economic entities covering the whole cycle of production, storage, processing, distribution, and consumption of food products, united and interacting based on economic relations and contributing to the achievement of individual and aggregate system effects. The study presents a structural scheme of integration and interaction of agrifood system’s elements, which embrace economic entities from agriculture and food sector, vertically integrated companies, households. In addition, it substantiates a conceptual approach to systematising factors of forecasting into endogenous and exogenous. The latter are divided into factors of the first and the second level in relation to economic entities depending on the degree of the influence. The analysis of particularities of exogenous factors’ manifestation in Russia’s agrifood system functioning reveals its instability originating from high volatility inherent in these factors, which confirms the pertinence of constructing predictive models based on neural networks. The findings allow for a broader understanding of the impact various factors produce on the parameters of agrifood system functioning, and can underlie the development of the relevant methods for forecasting using neural networks.