PREPARATION OF RETROSPECTIVE DATA FOR A NEURAL NETWORK MANAGEMENT SYSTEM FOR PROGRAMMED AGRICULTURAL PRODUCTION IN ARID CONDITIONS OF THE VOLGOGRAD REGION

Abstract

The article deals with methodological and applied issues of developing a database containing information for a neural network management system for programmed agricultural production on the example of long-term retrospective data on crop yields in arid conditions. Crop productivity is determined by a combination of various groups of endogenous and exogenous factors that take into account regional weather and climate conditions, physical, mechanical and agrochemical characteristics of soils and features of agricultural practices and technologies, including reclamation data that should be accumulated and stored in the database being developed. For the arid conditions of the Lower Volga region, including the Volgograd region, an important specific factor of crop variation is the impact of adverse natural phenomena (droughts, dry spells, frosts, etc.). The development of an agricultural production management system based on an artificial neural network (ins) requires the creation of specialized databases, the structure of which is determined by the tasks of agricultural production, including irrigation reclamation, necessary for the cultivation of crops in the arid conditions of the Lower Volga region. Justification of the structure and preparation of retrospective data for the creation and training of ins focused on the management of programmable agricultural production is an important scientific, methodological and applied task. In accordance with this, the database being developed for neural network management of agricultural production based on long-term retrospective data should include information blocks that differ in types and temporal characteristics of the accumulated information. The database can be pre-formed in the universal MS Excel format with subsequent conversion to csv format for in-depth statistical processing using the Numpy and Matplotlib mathematical libraries of the Python environment, where software prototypes of the developed ins are built.