Building a long-term forecast of grain production for risk management tasks

Abstract

Subject: forecasting the productivity indicators of grain production is an interdisciplinary problem that climatologists, agrometeorologists, mathematicians, agronomists and specialists in other fields are actively engaged in solving. A qualitative forecast of the productivity of grain production and scenarios of its development dynamics will allow the LPR (decision maker) to control and regulate the ambitious plans of Russian producers, spelled out in the Longterm strategy for the development of the grain complex of the Russian Federation until 2035. Purpose: construction of a predictive model of the productivity index of grain production in the Stavropol Territory based on the mechanism of operation of a linear cel- lular automaton. Research design: under the assumption that forecasting the productivity of grain production develops along cyclic trajectories, the stability of the characteristics of which is significantly higher than the stability of the periodicity of separately selected process points. The paper presents the concepts of "depth of memory", "longterm memory", as well as a description of the artificial intelligence method, its approbation and interpretation of the results obtained. Results: the authors present a demonstration of the operation of the linear cellular automaton method based on the time series of grain yields of the Stavropol Territory for the period from 1956-2020. The results of empirical studies have confirmed the possibility of practical use of the developed predictive models to justify management decisions.