Optimizing crop prediction, yield prediction and fertilizer recommendation with machine learning, feature selection, and sampling techniques for sustainable agriculture

For modern farmers, predicting crop yields is essential to making data-driven choices that increase productivity and promote sustainable farming practices. In order to predict crop yields based on a number of crucial soil and environmental characteristics, this study will use the machine learning approach. Soil nutrient levels, weather, temperature, and precipitation are all factors that go into training the model. This study employs a stable prediction model that is built using the following machine learning algorithms: Random Forest, Decision Tree, Naive Bayes, KNN, and Support Vector Machine (SVM). It is suggested to preprocess the data using feature selection techniques like BORUTA and Recursive Feature Elimination (RFE) to remove multiple or unnecessary features. This improves the accuracy and efficiency of the model. The database was prepared and the model accuracy was improved using random oversampling (ROSE) and SMOTE methods. The system is also used to provide a fertilizer recommendation whereas the system uses information on the soil and crop types and suggests effective fertilizers to use hereby helping farmers choose the best fertilizers in terms of nutrient management.