An Explainable Hybrid Learning Framework for Crop Recommendation

Choosing a suitable crop with respect to soil and climate conditions is one of the essential activities in precision agriculture. However, traditional machine learning models give reliable prediction but fail to interpret the results, which leads to untrusted automation recommendations. In this paper, an eXplainable Hybrid Learning Framework (EHLF) is developed by combining Random Forest (RF), eXtreme Gradient Boosting (XGBoost), and Multilayer Perceptron (MLP) using weighted soft voting for recommending the most suitable crops. EHLF uses seven agricultural attributes, which include nitrogen (N), phosphorus (P), potassium (K), temperature, humidity, pH, and rainfall to recommend the suitable crop. In order to increase the explanation ability of the framework, SHapley Additive exPlanations (SHAP) is used to explain the impact of each attribute on the prediction made by the framework.