Smart Crop Recommendation System Using Machine Learning and Explainable AI

In many countries of the world, especially in South Asian countries, most of the people's livelihood and food habits are based on agriculture, so it is very important to produce different types of crops profitably and in large quantities. However many farmers are still using primitive methods of cropping and do not understand exactly which crop is best for which land at which time due to climate variability, soil degradation, and environmental instability. The AI-based automated system is a crying need for the farmers for this problem. This study proposes a smart crop recommendation model using Machine Learning (ML) and Explainable AI to help farmers make data-driven decisions on what crops are ideal to grow. This investigation utilizes a standard crop recommendation dataset from Kaggle, which includes key soil characteristics such as nitrogen, phosphorus, potassium, and pH level; and climate parameters including temperature, humidity, and rainfall. The proposed work makes the best use of several machine learning algorithms including Support Vector Machine, Decision Tree, Random Forest, Gaussian Naive Bayes, and a boosting ensemble classifier called Adaptive Boosting (AdaBoost). Among these models, AdaBoost reveals the best performance of 99.35% with a Decision Tree oriented base learner ID3. In addition, the SHapley Additive exPlanations (SHAP) and Local Interpretable Model-agnostic Explanations (LIME) from explainable AI are used to detect the responsible soil and climate features for the classification of the AdaBoost ensemble model.