Developing an XAI-Based Crop Recommendation Framework Using Soil Nutrient Profiles and Historical Crop Yields

The agriculture sector is a significant cornerstone of the country and has boomed drastically in past decade, due to the advent of cutting-edge technologies such as Artificial Intelligence (AI), Machine Learning (ML), Blockchain, Cloud Computing, and Internet of Things (IoT). Recent studies have utilized ML models to overcome challenges like variability in climate patterns, soil fertility depletion, water scarcity, market volatility and optimal crop recommendation, by considering the features such as soil nutrients, climate conditions, rainfall, temperature, and fertilizers that significantly transform the agricultural sector. However, transparency, interpretability, and accountability of underlying models’ decisions are often neglected in real-life use cases. Hence, the authors have proposed a hyperparameter optimization-based grid search algorithm for smart agriculture framework that recommends the optimum crop with high accuracy utilizing the vital features. The proposed framework offers clear explanations, and transparency for each crop recommended to the farmers and demonstrates the key factors influencing the decision-making process using eXplainable-AI (XAI). Furthermore, the study is compared with conventional ML models such as Logistic Regression (LR), Decision Tree Classifier (DTC), Random Forest Classifier (RFC), K-Nearest Neighbor (KNN), Support Vector Machine (SVM), and Gradient Boosting (GB) to measure the performance. Consequently, the hyperparameter optimization-based grid search algorithm outperforms other state-of-the-art approaches in terms of accuracy up to 99.73%, precision by 95.05%, RMSE value by 99.52%, MAE value by 99.36% and R2 value by 99.17 thereby enhancing the validation and transparency of crop predictions by potential integration of XAI-based approaches.