Evaluation of Some Supervised Machine Learning Techniques for the Prediction of Soil Macro-Nutrients for Cash Crop Production

The prediction of soil macro-nutrients level is critical for optimizing cash crop production, ensuring both economic viability and sustainable agricultural practices.Researchers have used several machine learning models to predict the nutrients for the good yield of the crops; however, the supply and demand based on nutrients that provide the good yield cannot be met.Based on this shortfall, this study aims to evaluate some machine learning techniques for predicting soil nutrient for cash crop production.The dataset was sourced from "Nigeria Soils Data" on Africa Geoportal and includes soil samples collected from various locations across Nigeria.The data were preprocessed to handle missing values, feature engineering to transform spectral data using Principal Component Analysis (PCA), normalization of data features, and the splitting of the dataset.Each model was trained on the preprocessed data and assessed using Mean Absolute Error (MAE), Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and R score.The performance evaluation results for MAE, MSE, R , RMSE under Nitrogen are: 0.029078, 0.002053, 0.710939, 0.045315 respectively.The result emphasizes the superior performance of Random Forest (RF) as it outperforms the remaining models within the metrics of MAE, MSE, R , RMSE even after employing approaches to improve individual model performance through bagging methods.These insights can help agricultural stakeholders determine which approaches to employ, leading to enhance crop production and promote more eco-friendly agricultural methods.