Forecasting the tax burden of agricultural enterprises by machine learning methods

Abstract

The article analyzes the data of a set of agricultural enterprises and builds machine learning models to predict the tax burden. The subject of this study is a system of statistical indicators of agricultural enterprises that characterize the level of tax burden. The purpose of the study is to predict the tax burden using machine learning methods. The introduction of modern artificial intelligence tools is an integral and inevitable process in all spheres, including in the tax environment. Machine learning methods were used to build models: regression analysis, decision tree, random forest, gradient boosting. Models of forecasting the tax burden depending on a set of factors were built. The high quality of tax burden forecasting models will make it possible to more accurately assess the financial condition of enterprises, calculate profitability, predict profitability and make informed investment management decisions. As a result of forecasting the tax burden, the gradient boosting machine learning model turned out to have the best quality. In general, the model allows you to predict the tax burden better than traditional econometric models and make high-quality forecasts. The introduction of modern forecasting tools based on artificial intelligence methods will allow obtaining highly accurate forecasts with minimal time, which will increase the efficiency of enterprises and the level of production.