Crop recommendation system by comparing and analysing yield predictions

Abstract

In an era of rapid agricultural development, accurately predicting crop yield and recommending the most suitable crops is crucial for improving productivity and sustainability. This research focuses on building a robust Crop Recommendation System by analyzing yield predictions through the application of advanced machine learning algorithms and explainable AI (XAI) techniques. Utilizing datasets comprising soil properties, climatic conditions, and historical crop yield data, this study employs Decision Trees (DT), Random Forests (RF), Gradient Boosting Machines (XGBoost), AdaBoost, and Gaussian Naive Bayes (GNB). Furthermore, a Stacking Classifier integrates multiple base models to enhance prediction accuracy. To provide interpretability, Shapley Additive Explanations (SHAP) and Local Interpretable Model-agnostic Explanations (LIME) are incorporated, enabling stakeholders to understand the impact of individual features on model predictions. By comparing the performance and feature importance rankings across algorithms, the system identifies the optimal approach for yield prediction and crop recommendation. This analysis not only improves prediction reliability but also fosters data-driven agricultural decision-making, ultimately benefiting farmers and policymakers by maximizing yields, optimizing resource usage, and promoting sustainable agricultural practices.

Keywords: Crop recommendation, yield prediction, machine learning, explainable AI, SHAP, LIME, Decision Tree, Random Forest, XGBoost, AdaBoost, Gaussian Naive Bayes, Stacking Classifier