Predictive fertilization models for potato crops using machine learning techniques in Moroccan Gharb region

Abstract

Given the influence of several factors, including weather, soils, land management, genotypes, and the severity of pests and diseases, prescribing adequate nutrient levels is difficult. A potato’s performance can be predicted using machine learning techniques in cases when there is enough data. This study aimed to develop a highly precise model for determining the optimal levels of nitrogen, phosphorus, and potassium required to achieve both high-quality and high-yield potato crops, taking into account the impact of various environmental factors such as weather, soil type, and land management practices. We used 900 field experiments from Kaggle as part of a data set. We developed, evaluated, and compared prediction models of k-nearest neighbor (KNN), linear support vector machine (SVM), naive Bayes (NB) classifier, decision tree (DT) regressor, random forest (RF) regressor, and eXtreme gradient boosting (XGBoost). We used measures such as mean average error (MAE), mean squared error (MSE), R-Squared (RS), and R 2 Root mean squared error (RMSE) to describe the model’s mistakes and prediction capacity. It turned out that the XGBoost model has the greatest R 2, MSE and MAE values. Overall, the XGBoost model outperforms the other machine learning models. In the end, we suggested a hardware implementation to help farmers in the field.