Machine learning methods in soil fertility forecasting

Abstract

Listen

Translate article

The study presents the conclusions based on annotations of foreign articles in order to summarize the experience of using various machine learning technologies to solve a list of tasks for predicting yields on various scales and using heterogeneous data. We tested some of the techniques used in scientific circles on the collected data set with a strong focus on soils as a comprehensive indicator of yield. During the experiments with RF and XGB regression models, the hypothesis of the critical importance of soils was not justified. In the course of the study, the authors performed the works that have practical significance for agricultural enterprises.