Machine learning in Time Series Forecasting Taking into Account the Influence of External Factors in the Agro-Industrial Area.

Abstract

The work is devoted to the development and application of time series forecasting methods in agriculture, taking into account external factors. Modern approaches based on deep learning (Neural Prophet) and gradient boosting (Cat Boost) are analyzed to improve the accuracy of forecasting prices for agricultural products. The study includes an analysis of the impact of various factors such as feed costs, government regulations, news and seasonality, as well as the development of a strategy for encoding categorical features (One-Hot Encoding, Ordinal Encoding). The results demonstrate a significant improvement in the accuracy of forecasts compared to traditional methods, emphasizing the importance of integrating external data for making effective management decisions in the agroindustrial complex. doi 10.54708/19926502_2024_28410644