Application of Machine Learning Methods for Forecasting in the Agricultural Sector of the Economy

Abstract

This article is aimed at researching machine learning methods in the context of agriculture. Studying the issues of advanced mathematical method adaptation for the agricultural sector of the economy allows us to build effective management models and predict further development. The main purpose of this article is to analyze the application of advanced machine learning methods (linear regression, gradient boosting, random forest, neural networks (Deep Learning), Support Vector Machines (SVR), One-Class SVM, KNN, K-Means, principal component analysis (PCA)) in the agricultural sector in order to optimize production, increase the efficiency of agribusiness and predict parameters. Based on the analysis of key machine learning methods, conclusions regarding their applicability in agriculturewere drawn. A special emphasis in the study is placed on the development and improvement of a model for predicting agricultural yields using machine learning methods. For practical testing, a basic model based on RandomForestRegressor was chosen, which allows predicting yield, taking into account various factors such as weather conditions, fertilizer level, soil type and other significant aspects. As part of the study, an algorithm for performing calculations using Python is presented, which allows for model training, visualization of the influence of parameters and statistical analysis. During the testing process, the model was optimized using GridSearchCV, aimed at selecting optimal hyperparameters. The analysis revealed the possibility of model improvement, yet, at low statistical significance. The obtained research results show that it is possible to apply machine learning to improve the productivity of forecasting key parameters in agricultureeffectively.