Forecasting The Fish And Seafood Catch In The Global Economy

The article proposes a new scientific approach to forecasting the catch of fish and seafood, based on the use of the mathematical apparatus of neural networks (namely, Kohonen self-organizing maps) and econometric models (autoregressive functions). Self-organizing maps provided an opportunity to cluster the countries of the world according to the indicators of fish and seafood catch, which made it possible to identify countries and areas whose state and development of the fisheries sector are similar to each other. Due to the too heterogeneous state of this industry in different countries, the clustering procedure had to be repeated even within certain groups of countries. After obtaining a homogeneous sampling to solve the problem of predicting the volume of fish and seafood catches, it is proposed to apply time series regression models. Through a series of experiments, fifth-order autoregression AR(5) was determined as the best predictive model. The forecast obtained and model quality indicators approve the feasibility of applying the proposed multi-level approach to building predictive models, which is based on the clustering of countries of the world and the formation of homogeneous training samples. A similar approach can be used in a wide range of economic and managerial tasks.