Predicting macroelement content in legumes with machine learning

This study aims to develop accurate and efficient machine learning models to predict the concentrations of phosphorus (P), potassium (K), calcium (Ca), and magnesium (Mg) in 10 legume species naturally growing in the Çamlıhemşin district of Rize province, Türkiye. A comprehensive dataset of feed quality characteristics was collected, and four widely used machine learning algorithms-Multivariate Adaptive Regression Splines (MARS), K-Nearest Neighbors (KNN), Support Vector Regression (SVR), and Artificial Neural Networks (ANN)-were employed to build predictive models. The performance of these models was evaluated using a range of statistical metrics, including root mean squared error (RMSE), mean absolute error (MAE), and coefficient of determination (R²). Results indicated that the MARS model generally outperformed the others, achieving the lowest RMSE values and relatively high R² values for most elements, suggesting it is the most suitable model for predicting macroelement content in this particular dataset. KNN showed reasonable performance, while SVR and ANN exhibited relatively poor results, likely due to the limited dataset size and their sensitivity to hyperparameter settings. The study contributes to the advancement of precision agriculture by providing a robust and accurate method for assessing the nutritional quality of legume species.