AI and Machine Learning for Optimal Crop Yield Optimization in the USA

Abstract

The agricultural sector plays a paramount role in the economy of the United States, contributing significantly to the GDP and affirming sustainability for American residents. This study explored the application of Artificial Intelligence and Machine Learning techniques in maximizing crop yields in America. This research employed various software tools, comprising Python programming language, Pandas library for data manipulation and analysis, Scikit-learn library for machine learning models and evaluation metrics, and LIME library for explainable AI. The crop yield datasets for the current research were sourced from Kaggle. This dataset provided substantial insights regarding crop cultivation practices within the USA context. This study proposes the "XAI-CROP" algorithm, which is a novel explainable artificial intelligence (XAI) model developed particularly to reinforce the interpretability, transparency and trustworthiness of crop recommendation systems (CRS). From the experimentation, the XAI-CROP model excelled at forecasting crop yield, as demonstrated by its lowest MSE value of 0.9412, suggesting minimal errors.  Besides, Its MAE of 0.9874 suggests an average error of less than 1 unit in forecasting crop yield. Furthermore, the R2 value of 0.94152 suggests that the algorithm accounts for 94.15% of the data's variability.