Artificial Intelligence in Agricultural Management: Use of Random Forest Models for the Prediction of Seed Production and Reservation in Brazil

Abstract

This study addresses the application of Artificial Intelligence (AI) models, more specifically random forest, for the prediction of seed production and reserve in Brazilian agriculture. The main objective is to contribute to the advancement of resource management and planning, a critical action to increase the efficiency and sustainability of the sector. The work is highlighted by the importance of understanding the role of AI in optimizing agricultural practices, providing a framework for future research at the intersection between AI technologies and agriculture. Methodologically, the study implemented a rigorous data collection and processing process provided by the Ministry of Agriculture and Livestock of Brazil, covering the harvests from 2016/2017 to 2023/2024. Data cleansing preceded the transformation of categorical variables through one-hot coding and subsequent splitting of the dataset into 80% for training and 20% for testing. Using the scikit-learn library, a random forest model was configured and evaluated, employing validation techniques such as training/test split and cross-validation, in addition to mean square error (MSE) and coefficient of determination (R²) metrics to measure the accuracy and effectiveness of the model. The results indicate a moderate to strong positive correlation between the variables of time and number of seeds reserved for both growing periods, Safra and Safrinha. However, the analyses pointed to annual variability and differentiated confidence in predictions between periods, suggesting the influence of additional factors and the need for adaptive models. The concentration of production in a few cultures was identified as a potential risk, suggesting that diversification is key to the resilience of the sector. The generalizability of the model was evaluated, and the phenomenon of overfitting was considered a possibility given the variations in accuracy between the training and test data. This study reinforces the transformative potential that AI models, such as the random forest, possess for agricultural prediction and management, opening doors for future improvements and providing valuable subsidies for data-driven strategic decisions in the agricultural sector.

DOI:https://doi.org/10.56238/sevened2024.023-006