Forecasting the production of gross output in agricultural sector of the Ryazan oblast

Abstract

Gross agricultural output is a generalised physical output indicator in an industry that includes hundreds of different types of products, as well as the result of the interaction of production factors. This study provides a comparative analysis of methods based on "Gradient boosting of regression trees" in the Python programming language to identify the optimal values of the model parameters with the subsequent construction of a predictive model based on indicators that affect the production of gross agricultural output. The purpose of this study is to build a regression model for predicting gross agricultural output at actual prices for 2020. To achieve this goal, the methods of regression analysis, forecasting, gradient boosting, etc., were used. The gradient boosting of regression trees was solved for the conditions of the Ryazan Oblast. 4 models were created, 2 of which were based on the preliminary data processing. As a result of the construction of all models, the optimal values of the parameters were found and the results of the correctness on the model on the test set were obtained. It was found that the gradient boosting of regression trees gives adequate regression models for predicting the target variable, in particular, the indicator of gross agricultural output. The investigated indicator is a complex result of the interaction of many factors that are common for agricultural production. Thus, the gradient boosting of trees is most suitable for forecasting complex open systems. Such a method can be used to forecast the production of gross agricultural output not only of individual regions but also of the state as a whole. Based on the "test_score" model, which showed the correctness of 99% (0.994) on the test set, the gross agricultural output in all categories of farms in 2020 amounted to RUB 19187.84 million.